[AMDGPU] Vectorize more 16 bit shuffles (#90648)

In the case of larger vectors, we should still prefer the vectorized
version (i.e. shufflevector vs extract/insert chains).

In arithmetic chains, vectorization results in chains of packed math
instructions (as opposed to unpack/repack & scalarized arithmetic):
https://godbolt.org/z/c5onaf6G5

In chains with PHIs, vectorization again removes the unnecessary pack /
repack code around BBs: https://godbolt.org/z/vz7zYzvhs

7 files changed, 1278 insertions, 724 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
index 84320d2..437e01c 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -1129,31 +1129,56 @@ InstructionCost GCNTTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
                                            int Index, VectorType *SubTp,
                                            ArrayRef<const Value *> Args,
                                            const Instruction *CxtI) {
+  if (!isa<FixedVectorType>(VT))
+    return BaseT::getShuffleCost(Kind, VT, Mask, CostKind, Index, SubTp);
+
   Kind = improveShuffleKindFromMask(Kind, Mask, VT, Index, SubTp);
-  // Treat extractsubvector as single op permutation.
-  bool IsExtractSubvector = Kind == TTI::SK_ExtractSubvector;
-  if (IsExtractSubvector)
-    Kind = TTI::SK_PermuteSingleSrc;
-
-  if (ST->hasVOP3PInsts()) {
-    if (cast<FixedVectorType>(VT)->getNumElements() == 2 &&
-        DL.getTypeSizeInBits(VT->getElementType()) == 16) {
-      // With op_sel VOP3P instructions freely can access the low half or high
-      // half of a register, so any swizzle is free.
 
-      switch (Kind) {
-      case TTI::SK_Broadcast:
-      case TTI::SK_Reverse:
-      case TTI::SK_PermuteSingleSrc:
+  // Larger vector widths may require additional instructions, but are
+  // typically cheaper than scalarized versions.
+  unsigned NumVectorElts = cast<FixedVectorType>(VT)->getNumElements();
+  if (ST->getGeneration() >= AMDGPUSubtarget::VOLCANIC_ISLANDS &&
+      DL.getTypeSizeInBits(VT->getElementType()) == 16) {
+    bool HasVOP3P = ST->hasVOP3PInsts();
+    unsigned RequestedElts =
+        count_if(Mask, [](int MaskElt) { return MaskElt != -1; });
+    if (RequestedElts == 0)
+      return 0;
+    switch (Kind) {
+    case TTI::SK_Broadcast:
+    case TTI::SK_Reverse:
+    case TTI::SK_PermuteSingleSrc: {
+      // With op_sel VOP3P instructions freely can access the low half or high
+      // half of a register, so any swizzle of two elements is free.
+      if (HasVOP3P && NumVectorElts == 2)
         return 0;
-      default:
-        break;
-      }
+      unsigned NumPerms = alignTo(RequestedElts, 2) / 2;
+      // SK_Broadcast just reuses the same mask
+      unsigned NumPermMasks = Kind == TTI::SK_Broadcast ? 1 : NumPerms;
+      return NumPerms + NumPermMasks;
+    }
+    case TTI::SK_ExtractSubvector:
+    case TTI::SK_InsertSubvector: {
+      // Even aligned accesses are free
+      if (!(Index % 2))
+        return 0;
+      // Insert/extract subvectors only require shifts / extract code to get the
+      // relevant bits
+      return alignTo(RequestedElts, 2) / 2;
+    }
+    case TTI::SK_PermuteTwoSrc:
+    case TTI::SK_Splice:
+    case TTI::SK_Select: {
+      unsigned NumPerms = alignTo(RequestedElts, 2) / 2;
+      // SK_Select just reuses the same mask
+      unsigned NumPermMasks = Kind == TTI::SK_Select ? 1 : NumPerms;
+      return NumPerms + NumPermMasks;
+    }
+
+    default:
+      break;
     }
   }
-  // Restore optimal kind.
-  if (IsExtractSubvector)
-    Kind = TTI::SK_ExtractSubvector;
 
   return BaseT::getShuffleCost(Kind, VT, Mask, CostKind, Index, SubTp);
 }
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll b/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
index be5cca0..a181567 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/shufflevector.ll
@@ -7,603 +7,1140 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=amdgcn-unknown-amdhsa -mcpu=fiji -cost-kind=code-size -S | FileCheck -check-prefixes=ALL-SIZE,VI-SIZE %s
 ; END.
 
-define amdgpu_kernel void @shufflevector_i16() {
+define amdgpu_kernel void @shufflevector_i16(<2 x i16> %vec1, <2 x i16> %vec2) {
 ; GFX9-10-LABEL: 'shufflevector_i16'
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> zeroinitializer
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 3>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf000 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> zeroinitializer
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf222 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 2>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; VI-LABEL: 'shufflevector_i16'
-; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> zeroinitializer
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 3>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf000 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> zeroinitializer
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf222 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 2>
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX9-10-SIZE-LABEL: 'shufflevector_i16'
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> zeroinitializer
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 3>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf000 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> zeroinitializer
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf222 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 2>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; VI-SIZE-LABEL: 'shufflevector_i16'
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf00 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> zeroinitializer
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf22 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 3>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf000 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> zeroinitializer
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf222 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 2>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> zeroinitializer
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> zeroinitializer
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
-  %shuf00 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> zeroinitializer
-  %shuf01 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 1>
-  %shuf10 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-  %shuf11 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-  %shuf02 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 2>
-  %shuf20 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 0>
-  %shuf22 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 2>
-  %shuf03 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 3>
-  %shuf30 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 0>
-  %shuf33 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 3>
-  %shuf12 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 2>
-  %shuf21 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 1>
-  %shuf13 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 3>
-  %shuf31 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 1>
-  %shuf23 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 2, i32 3>
-  %shuf32 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 3, i32 2>
-  %shuf000 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 0>
-  %shuf001 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 1>
-  %shuf010 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 0>
-  %shuf011 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 1, i32 1>
-  %shuf100 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 0>
-  %shuf101 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 0, i32 1>
-  %shuf110 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 0>
-  %shuf111 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 1>
-  %shuf002 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 0, i32 2>
-  %shuf020 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 0>
-  %shuf022 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 0, i32 2, i32 2>
-  %shuf200 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 0>
-  %shuf202 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 0, i32 2>
-  %shuf220 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 0>
-  %shuf222 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 2>
-  %shuf112 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 1, i32 2>
-  %shuf121 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 1>
-  %shuf122 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 1, i32 2, i32 2>
-  %shuf211 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 1>
-  %shuf212 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 1, i32 2>
-  %shuf221 = shufflevector <2 x i16> undef, <2 x i16> undef, <3 x i32> <i32 2, i32 2, i32 1>
+  %shuf00 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> zeroinitializer
+  %shuf01 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 1>
+  %shuf10 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 0>
+  %shuf11 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 1>
+  %shuf02 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 2>
+  %shuf20 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 0>
+  %shuf22 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 2>
+  %shuf03 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 0, i32 3>
+  %shuf30 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 0>
+  %shuf33 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 3>
+  %shuf12 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 2>
+  %shuf21 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 1>
+  %shuf13 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 1, i32 3>
+  %shuf31 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 1>
+  %shuf23 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 2, i32 3>
+  %shuf32 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <2 x i32> <i32 3, i32 2>
+  %shuf000 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 0>
+  %shuf001 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+  %shuf010 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+  %shuf011 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+  %shuf100 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+  %shuf101 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+  %shuf110 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+  %shuf111 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+  %shuf002 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+  %shuf020 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+  %shuf022 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+  %shuf200 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+  %shuf202 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+  %shuf220 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+  %shuf222 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+  %shuf112 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+  %shuf121 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+  %shuf122 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+  %shuf211 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+  %shuf212 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+  %shuf221 = shufflevector <2 x i16> %vec1, <2 x i16> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+  %shuf00_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> zeroinitializer
+  %shuf01_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 1>
+  %shuf10_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 0>
+  %shuf11_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 1>
+  %shuf02_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 2>
+  %shuf20_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 0>
+  %shuf22_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 2>
+  %shuf03_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 0, i32 3>
+  %shuf30_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 0>
+  %shuf33_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 3>
+  %shuf12_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 2>
+  %shuf21_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 1>
+  %shuf13_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 1, i32 3>
+  %shuf31_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 1>
+  %shuf23_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 2, i32 3>
+  %shuf32_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <2 x i32> <i32 3, i32 2>
+  %shuf000_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 0>
+  %shuf001_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+  %shuf010_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+  %shuf011_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+  %shuf100_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+  %shuf101_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+  %shuf110_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+  %shuf111_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+  %shuf002_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+  %shuf020_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+  %shuf022_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+  %shuf200_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+  %shuf202_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+  %shuf220_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+  %shuf222_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+  %shuf112_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+  %shuf121_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+  %shuf122_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+  %shuf211_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+  %shuf212_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+  %shuf221_2 = shufflevector <2 x i16> %vec1, <2 x i16> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
   ret void
 }
 
 ; Should not assert
-define amdgpu_kernel void @shufflevector_i8() {
+define amdgpu_kernel void @shufflevector_i8(<2 x i8> %vec1, <2 x i8> %vec2) {
 ; ALL-LABEL: 'shufflevector_i8'
-; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf00 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> zeroinitializer
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf10 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf11 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf20 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf22 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf03 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf30 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf33 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf12 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf21 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf31 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf32 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> zeroinitializer
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf001 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf010 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf011 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf100 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 0, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf101 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf110 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf111 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf002 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf020 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf022 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf200 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 0, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf202 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf220 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf112 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf121 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf122 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf211 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf212 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf221 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf00 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf10 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf11 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf20 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf22 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf03 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf30 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf33 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf12 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf21 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf31 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf32 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf000 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf001 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf010 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf011 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf100 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf101 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf110 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf111 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf002 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf020 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf022 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf200 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf202 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf220 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf222 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf112 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf121 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf122 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf211 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf212 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf221 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf00_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf10_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf11_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf20_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf22_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf03_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf30_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf33_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf12_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf21_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf31_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf32_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf000_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf001_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf010_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf011_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf100_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf101_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf110_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf111_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf002_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf020_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf022_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf200_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf202_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf220_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf222_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf112_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf121_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf122_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf211_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf212_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf221_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'shufflevector_i8'
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf00 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> zeroinitializer
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf10 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf11 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf20 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf22 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf03 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf30 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf33 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf12 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf21 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf31 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf32 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf000 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> zeroinitializer
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf001 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf010 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf011 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf100 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 0, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf101 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf110 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf111 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf002 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf020 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf022 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf200 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 0, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf202 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf220 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf222 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf112 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf121 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf122 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf211 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf212 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf221 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf00 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf10 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf11 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf20 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf22 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf03 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf30 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf33 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf12 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf21 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf31 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf32 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf000 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf001 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf010 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf011 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf100 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf101 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf110 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf111 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf002 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf020 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf022 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf200 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf202 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf220 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf222 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf112 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf121 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf122 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf211 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf212 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf221 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf00_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf10_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf11_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf02_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shuf20_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shuf22_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf03_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf30_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf33_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf12_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf21_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf13_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf31_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shuf32_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf000_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf001_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf010_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf011_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf100_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf101_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf110_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf111_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf002_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf020_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf022_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf200_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf202_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf220_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf222_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf112_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf121_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf122_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf211_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf212_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shuf221_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
-  %shuf00 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> zeroinitializer
-  %shuf01 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 1>
-  %shuf10 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-  %shuf11 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-  %shuf02 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 2>
-  %shuf20 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 0>
-  %shuf22 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 2>
-  %shuf03 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 3>
-  %shuf30 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 0>
-  %shuf33 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 3>
-  %shuf12 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 2>
-  %shuf21 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 1>
-  %shuf13 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 3>
-  %shuf31 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 1>
-  %shuf23 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 2, i32 3>
-  %shuf32 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 3, i32 2>
-  %shuf000 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 0, i32 0>
-  %shuf001 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 0, i32 1>
-  %shuf010 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 1, i32 0>
-  %shuf011 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 1, i32 1>
-  %shuf100 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 0, i32 0>
-  %shuf101 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 0, i32 1>
-  %shuf110 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 0>
-  %shuf111 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 1>
-  %shuf002 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 0, i32 2>
-  %shuf020 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 2, i32 0>
-  %shuf022 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 0, i32 2, i32 2>
-  %shuf200 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 0, i32 0>
-  %shuf202 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 0, i32 2>
-  %shuf220 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 0>
-  %shuf222 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 2>
-  %shuf112 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 1, i32 2>
-  %shuf121 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 2, i32 1>
-  %shuf122 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 1, i32 2, i32 2>
-  %shuf211 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 1, i32 1>
-  %shuf212 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 1, i32 2>
-  %shuf221 = shufflevector <2 x i8> undef, <2 x i8> undef, <3 x i32> <i32 2, i32 2, i32 1>
+  %shuf00 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> zeroinitializer
+  %shuf01 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 1>
+  %shuf10 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 0>
+  %shuf11 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 1>
+  %shuf02 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 2>
+  %shuf20 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 0>
+  %shuf22 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 2>
+  %shuf03 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 0, i32 3>
+  %shuf30 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 0>
+  %shuf33 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 3>
+  %shuf12 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 2>
+  %shuf21 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 1>
+  %shuf13 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 1, i32 3>
+  %shuf31 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 1>
+  %shuf23 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 2, i32 3>
+  %shuf32 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <2 x i32> <i32 3, i32 2>
+  %shuf000 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 0>
+  %shuf001 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+  %shuf010 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+  %shuf011 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+  %shuf100 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+  %shuf101 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+  %shuf110 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+  %shuf111 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+  %shuf002 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+  %shuf020 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+  %shuf022 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+  %shuf200 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+  %shuf202 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+  %shuf220 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+  %shuf222 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+  %shuf112 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+  %shuf121 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+  %shuf122 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+  %shuf211 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+  %shuf212 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+  %shuf221 = shufflevector <2 x i8> %vec1, <2 x i8> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+  %shuf00_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> zeroinitializer
+  %shuf01_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 1>
+  %shuf10_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 0>
+  %shuf11_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 1>
+  %shuf02_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 2>
+  %shuf20_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 0>
+  %shuf22_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 2>
+  %shuf03_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 0, i32 3>
+  %shuf30_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 0>
+  %shuf33_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 3>
+  %shuf12_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 2>
+  %shuf21_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 1>
+  %shuf13_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 1, i32 3>
+  %shuf31_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 1>
+  %shuf23_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 2, i32 3>
+  %shuf32_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <2 x i32> <i32 3, i32 2>
+  %shuf000_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 0>
+  %shuf001_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+  %shuf010_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+  %shuf011_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+  %shuf100_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+  %shuf101_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+  %shuf110_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+  %shuf111_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+  %shuf002_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+  %shuf020_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+  %shuf022_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+  %shuf200_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+  %shuf202_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+  %shuf220_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+  %shuf222_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+  %shuf112_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+  %shuf121_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+  %shuf122_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+  %shuf211_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+  %shuf212_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+  %shuf221_2 = shufflevector <2 x i8> %vec1, <2 x i8> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
   ret void
 }
 
-define amdgpu_kernel void @shufflevector_i32() {
+define amdgpu_kernel void @shufflevector_i32(<2 x i32> %vec1, <2 x i32> %vec2) {
 ; ALL-LABEL: 'shufflevector_i32'
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> zeroinitializer
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf02 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf20 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf30 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf12 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf13 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf31 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf000 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> zeroinitializer
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf001 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf011 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf100 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 0, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf101 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 0, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf110 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf111 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf002 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf020 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf022 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf200 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 0, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf202 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 0, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf220 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 0>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf222 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf112 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf121 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 2, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf122 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 2, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf211 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 1, i32 1>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf212 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 1, i32 2>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf221 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf02 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf20 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf30 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf12 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf13 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf31 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf000 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf001 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf011 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf100 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf101 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf110 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf111 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf002 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf020 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf022 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf200 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf202 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf220 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf222 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf112 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf121 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf122 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf211 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf212 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf221 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf02_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf20_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf30_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf12_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf13_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf31_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf000_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> zeroinitializer
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf001_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf011_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf100_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf101_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf110_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf111_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf002_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf020_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf022_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf200_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf202_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf220_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf222_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf112_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf121_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf122_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf211_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf212_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf221_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'shufflevector_i32'
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> zeroinitializer
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf02 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf20 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf30 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf12 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf13 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf31 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf000 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> zeroinitializer
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf001 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf011 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf100 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 0, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf101 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 0, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf110 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf111 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf002 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf020 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf022 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf200 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 0, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf202 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 0, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf220 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 0>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf222 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf112 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf121 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 2, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf122 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 2, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf211 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 1, i32 1>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf212 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 1, i32 2>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf221 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf02 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf20 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf30 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf12 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf13 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf31 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf000 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf001 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf011 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf100 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf101 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf110 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf111 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf002 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf020 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf022 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf200 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf202 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf220 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf222 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf112 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf121 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf122 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf211 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf212 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf221 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf00_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf01_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf10_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf11_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf02_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf20_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf22_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf03_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf30_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf33_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf12_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf21_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf13_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf31_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf23_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf32_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf000_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> zeroinitializer
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf001_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf010_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf011_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf100_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf101_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf110_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf111_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf002_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf020_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf022_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf200_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf202_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf220_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf222_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf112_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf121_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf122_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf211_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf212_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %shuf221_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
-  %shuf00 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> zeroinitializer
-  %shuf01 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 1>
-  %shuf10 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-  %shuf11 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-  %shuf02 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 2>
-  %shuf20 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 0>
-  %shuf22 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 2>
-  %shuf03 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 3>
-  %shuf30 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 0>
-  %shuf33 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 3>
-  %shuf12 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 2>
-  %shuf21 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 1>
-  %shuf13 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 3>
-  %shuf31 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 1>
-  %shuf23 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 2, i32 3>
-  %shuf32 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 3, i32 2>
-  %shuf000 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 0, i32 0>
-  %shuf001 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 0, i32 1>
-  %shuf010 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 1, i32 0>
-  %shuf011 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 1, i32 1>
-  %shuf100 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 0, i32 0>
-  %shuf101 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 0, i32 1>
-  %shuf110 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 0>
-  %shuf111 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 1>
-  %shuf002 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 0, i32 2>
-  %shuf020 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 2, i32 0>
-  %shuf022 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 0, i32 2, i32 2>
-  %shuf200 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 0, i32 0>
-  %shuf202 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 0, i32 2>
-  %shuf220 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 0>
-  %shuf222 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 2>
-  %shuf112 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 1, i32 2>
-  %shuf121 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 2, i32 1>
-  %shuf122 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 1, i32 2, i32 2>
-  %shuf211 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 1, i32 1>
-  %shuf212 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 1, i32 2>
-  %shuf221 = shufflevector <2 x i32> undef, <2 x i32> undef, <3 x i32> <i32 2, i32 2, i32 1>
+  %shuf00 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> zeroinitializer
+  %shuf01 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 1>
+  %shuf10 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 0>
+  %shuf11 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 1>
+  %shuf02 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 2>
+  %shuf20 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 0>
+  %shuf22 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 2>
+  %shuf03 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 0, i32 3>
+  %shuf30 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 0>
+  %shuf33 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 3>
+  %shuf12 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 2>
+  %shuf21 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 1>
+  %shuf13 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 1, i32 3>
+  %shuf31 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 1>
+  %shuf23 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 2, i32 3>
+  %shuf32 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <2 x i32> <i32 3, i32 2>
+  %shuf000 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 0, i32 0>
+  %shuf001 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 0, i32 1>
+  %shuf010 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 1, i32 0>
+  %shuf011 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 1, i32 1>
+  %shuf100 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 0, i32 0>
+  %shuf101 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 0, i32 1>
+  %shuf110 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 0>
+  %shuf111 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 1>
+  %shuf002 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 0, i32 2>
+  %shuf020 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 2, i32 0>
+  %shuf022 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 0, i32 2, i32 2>
+  %shuf200 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 0, i32 0>
+  %shuf202 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 0, i32 2>
+  %shuf220 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 0>
+  %shuf222 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 2>
+  %shuf112 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 1, i32 2>
+  %shuf121 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 2, i32 1>
+  %shuf122 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 1, i32 2, i32 2>
+  %shuf211 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 1, i32 1>
+  %shuf212 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 1, i32 2>
+  %shuf221 = shufflevector <2 x i32> %vec1, <2 x i32> %vec1, <3 x i32> <i32 2, i32 2, i32 1>
+  %shuf00_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> zeroinitializer
+  %shuf01_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 1>
+  %shuf10_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 0>
+  %shuf11_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 1>
+  %shuf02_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 2>
+  %shuf20_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 0>
+  %shuf22_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 2>
+  %shuf03_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 0, i32 3>
+  %shuf30_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 0>
+  %shuf33_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 3>
+  %shuf12_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 2>
+  %shuf21_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 1>
+  %shuf13_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 1, i32 3>
+  %shuf31_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 1>
+  %shuf23_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 2, i32 3>
+  %shuf32_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <2 x i32> <i32 3, i32 2>
+  %shuf000_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 0, i32 0>
+  %shuf001_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 0, i32 1>
+  %shuf010_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 1, i32 0>
+  %shuf011_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 1, i32 1>
+  %shuf100_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 0, i32 0>
+  %shuf101_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 0, i32 1>
+  %shuf110_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 0>
+  %shuf111_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 1>
+  %shuf002_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 0, i32 2>
+  %shuf020_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 2, i32 0>
+  %shuf022_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 0, i32 2, i32 2>
+  %shuf200_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 0, i32 0>
+  %shuf202_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 0, i32 2>
+  %shuf220_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 0>
+  %shuf222_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 2>
+  %shuf112_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 1, i32 2>
+  %shuf121_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 2, i32 1>
+  %shuf122_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 1, i32 2, i32 2>
+  %shuf211_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 1, i32 1>
+  %shuf212_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 1, i32 2>
+  %shuf221_2 = shufflevector <2 x i32> %vec1, <2 x i32> %vec2, <3 x i32> <i32 2, i32 2, i32 1>
   ret void
 }
 
 ; Other shuffle cases
-define void @shuffle() {
+define void @shuffle(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8> %i8v4_2, <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i8> %i8v8, <8 x i8> %i8v8_2, <16 x i8> %i8v16, <16 x i8> %i8v16_2, <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i16> %i16v4, <4 x i16> %i16v4_2, <8 x i16> %i16v8, <8 x i16> %i16v8_2, <2 x i32> %i32v2, <2 x i32> %i32v2_2, <4 x i32> %i32v4, <4 x i32> %i32v4_2, <2 x float> %floatv2, <2 x float> %floatv2_2, <4 x float> %floatv4, <4 x float> %floatv4_2,<2 x i64> %i64v2, <2 x i64> %i64v2_2,<2 x double> %doublev2, <2 x double> %doublev2_2) {
 ; GFX9-10-LABEL: 'shuffle'
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8 = shufflevector <2 x i8> undef, <2 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8 = shufflevector <6 x i8> undef, <6 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v4i16_4 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v8i16_8 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4_2 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4_2 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2_2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; VI-LABEL: 'shuffle'
-; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8 = shufflevector <2 x i8> undef, <2 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8 = shufflevector <6 x i8> undef, <6 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v4i16_4 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v8i16_8 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4_2 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2_2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4_2 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2_2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <2 x i32> <i32 1, i32 0>
+; VI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2_2, <2 x i32> <i32 1, i32 0>
 ; VI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX9-10-SIZE-LABEL: 'shuffle'
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8 = shufflevector <2 x i8> undef, <2 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8 = shufflevector <6 x i8> undef, <6 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v4i16_4 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v8i16_8 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4_2 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4_2 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2_2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <2 x i32> <i32 1, i32 0>
+; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2_2, <2 x i32> <i32 1, i32 0>
 ; GFX9-10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; VI-SIZE-LABEL: 'shuffle'
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8 = shufflevector <2 x i8> undef, <2 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8 = shufflevector <6 x i8> undef, <6 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v4i16_4 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v8i16_8 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v2i8_8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_4_2 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f32_2_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2_2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_4_2 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2i64_2_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2_2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <2 x i32> <i32 1, i32 0>
+; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v2f64_2_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2_2, <2 x i32> <i32 1, i32 0>
 ; VI-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
-  %v2i8_2 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-  %v2i8_4 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-  %v4i8_4 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-  %v2i8_8 = shufflevector <2 x i8> undef, <2 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-  %v4i8_8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-  %v6i8_8 = shufflevector <6 x i8> undef, <6 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-  %v8i8_8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-  %v16i8_16 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-  %v2i16_2 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-  %v4i16_4 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-  %v8i16_8 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
-  %v2i32_2 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-  %v4i32_4 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-  %v2f32_2 = shufflevector <2 x float> undef, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-  %v4f32_4 = shufflevector <4 x float> undef, <4 x float> undef, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
-  %v2i64_2 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-  %v2f64_2 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 1, i32 0>
+  %v2i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <2 x i32> <i32 1, i32 0>
+  %v2i8_2_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <2 x i32> <i32 1, i32 0>
+  %v2i8_4 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v2i8_4_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v4i8_4 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v4i8_4_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v2i8_8 = shufflevector <2 x i8>  %i8v2, <2 x i8>  %i8v2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v2i8_8_2 = shufflevector <2 x i8>  %i8v2, <2 x i8>  %i8v2_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v4i8_8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v4i8_8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v6i8_8 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v6i8_8_2 = shufflevector <6 x i8> %i8v6, <6 x i8> %i8v6_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v8i8_8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v8i8_8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v16i8_16 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v16i8_16_2 = shufflevector <16 x i8> %i8v16, <16 x i8> %i8v16_2, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v2i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <2 x i32> <i32 1, i32 0>
+  %v2i16_2_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <2 x i32> <i32 1, i32 0>
+  %v4i16_4 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v4i16_4_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v8i16_8 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v8i16_8_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
+  %v2i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <2 x i32> <i32 1, i32 0>
+  %v2i32_2_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <2 x i32> <i32 1, i32 0>
+  %v4i32_4 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v4i32_4_2 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v2f32_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <2 x i32> <i32 1, i32 0>
+  %v2f32_2_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2_2, <2 x i32> <i32 1, i32 0>
+  %v4f32_4 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v4f32_4_2 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4_2, <4 x i32> <i32 1, i32 3, i32 2, i32 0>
+  %v2i64_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2, <2 x i32> <i32 1, i32 0>
+  %v2i64_2_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2_2, <2 x i32> <i32 1, i32 0>
+  %v2f64_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <2 x i32> <i32 1, i32 0>
+  %v2f64_2_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2_2, <2 x i32> <i32 1, i32 0>
   ret void
 }
 
-define void @concat() {
+define void @concat(<2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i8> %i8v8, <8 x i8> %i8v8_2, <2 x half> %halfv2, <2 x half> %halfv2_2, <4 x half> %halfv4, <4 x half> %halfv4_2, <8 x half> %halfv8, <8 x half> %halfv8_2, <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i16> %i16v4, <4 x i16> %i16v4_2, <8 x i16> %i16v8, <8 x i16> %i16v8_2, <2 x i32> %i32v2, <2 x i32> %i32v2_2, <4 x i32> %i32v4, <4 x i32> %i32v4_2, <2 x float> %floatv2, <2 x float> %floatv2_2, <4 x float> %floatv4, <4 x float> %floatv4_2,<2 x i64> %i64v2, <2 x i64> %i64v2_2,<2 x double> %doublev2, <2 x double> %doublev2_2) {
 ; ALL-LABEL: 'concat'
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i16 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i16 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f16 = shufflevector <2 x half> %halfv2, <2 x half> %halfv2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f16 = shufflevector <4 x half> %halfv4, <4 x half> %halfv4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16f16 = shufflevector <8 x half> %halfv8, <8 x half> %halfv8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i16_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32_2 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f16_2 = shufflevector <2 x half> %halfv2, <2 x half> %halfv2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f16_2 = shufflevector <4 x half> %halfv4, <4 x half> %halfv4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16f16_2 = shufflevector <8 x half> %halfv8, <8 x half> %halfv8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32_2 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'concat'
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i16 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i16 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f16 = shufflevector <2 x half> %halfv2, <2 x half> %halfv2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f16 = shufflevector <4 x half> %halfv4, <4 x half> %halfv4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16f16 = shufflevector <8 x half> %halfv8, <8 x half> %halfv8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i16_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16i16_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8i32_2 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4i64_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f16_2 = shufflevector <2 x half> %halfv2, <2 x half> %halfv2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f16_2 = shufflevector <4 x half> %halfv4, <4 x half> %halfv4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v16f16_2 = shufflevector <8 x half> %halfv8, <8 x half> %halfv8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f32_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v8f32_2 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %v4f64_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
-  %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-  %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-  %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-  %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-  %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-  %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-  %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-  %v8i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-  %v4i64 = shufflevector <2 x i64> undef, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-  %v4f16 = shufflevector <2 x half> undef, <2 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-  %v8f16 = shufflevector <4 x half> undef, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-  %v16f16 = shufflevector <8 x half> undef, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-  %v4f32 = shufflevector <2 x float> undef, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-  %v8f32 = shufflevector <4 x float> undef, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-  %v4f64 = shufflevector <2 x double> undef, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v4i8 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i8 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16i8 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4i16 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i16 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16i16 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4i32 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i32 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v4i64 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v4f16 = shufflevector <2 x half> %halfv2, <2 x half> %halfv2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8f16 = shufflevector <4 x half> %halfv4, <4 x half> %halfv4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16f16 = shufflevector <8 x half> %halfv8, <8 x half> %halfv8, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4f32 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8f32 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v4f64 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v4i8_2 = shufflevector <2 x i8> %i8v2, <2 x i8> %i8v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i8_2 = shufflevector <4 x i8> %i8v4, <4 x i8> %i8v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16i8_2 = shufflevector <8 x i8> %i8v8, <8 x i8> %i8v8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4i16_2 = shufflevector <2 x i16> %i16v2, <2 x i16> %i16v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i16_2 = shufflevector <4 x i16> %i16v4, <4 x i16> %i16v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16i16_2 = shufflevector <8 x i16> %i16v8, <8 x i16> %i16v8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4i32_2 = shufflevector <2 x i32> %i32v2, <2 x i32> %i32v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8i32_2 = shufflevector <4 x i32> %i32v4, <4 x i32> %i32v4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v4i64_2 = shufflevector <2 x i64> %i64v2, <2 x i64> %i64v2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v4f16_2 = shufflevector <2 x half> %halfv2, <2 x half> %halfv2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8f16_2 = shufflevector <4 x half> %halfv4, <4 x half> %halfv4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v16f16_2 = shufflevector <8 x half> %halfv8, <8 x half> %halfv8_2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %v4f32_2 = shufflevector <2 x float> %floatv2, <2 x float> %floatv2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %v8f32_2 = shufflevector <4 x float> %floatv4, <4 x float> %floatv4_2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %v4f64_2 = shufflevector <2 x double> %doublev2, <2 x double> %doublev2_2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   ret void
 }
diff --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat-inseltpoison.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat-inseltpoison.ll
index 2905601..3749bdf 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat-inseltpoison.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat-inseltpoison.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -passes=slp-vectorizer,instcombine %s | FileCheck -check-prefixes=GCN,GFX7 %s
 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -passes=slp-vectorizer,instcombine %s | FileCheck -check-prefixes=GCN,GFX8 %s
-; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -passes=slp-vectorizer,instcombine %s | FileCheck -check-prefixes=GCN,GFX8 %s
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -passes=slp-vectorizer,instcombine %s | FileCheck -check-prefixes=GCN,GFX9 %s
 
 define <2 x i16> @uadd_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX7-LABEL: @uadd_sat_v2i16(
@@ -21,6 +21,11 @@ define <2 x i16> @uadd_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX8-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
 ; GFX8-NEXT:    ret <2 x i16> [[TMP0]]
 ;
+; GFX9-LABEL: @uadd_sat_v2i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
+; GFX9-NEXT:    ret <2 x i16> [[TMP0]]
+;
 bb:
   %arg0.0 = extractelement <2 x i16> %arg0, i64 0
   %arg0.1 = extractelement <2 x i16> %arg0, i64 1
@@ -51,6 +56,11 @@ define <2 x i16> @usub_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX8-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.usub.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
 ; GFX8-NEXT:    ret <2 x i16> [[TMP0]]
 ;
+; GFX9-LABEL: @usub_sat_v2i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.usub.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
+; GFX9-NEXT:    ret <2 x i16> [[TMP0]]
+;
 bb:
   %arg0.0 = extractelement <2 x i16> %arg0, i64 0
   %arg0.1 = extractelement <2 x i16> %arg0, i64 1
@@ -81,6 +91,11 @@ define <2 x i16> @sadd_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX8-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.sadd.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
 ; GFX8-NEXT:    ret <2 x i16> [[TMP0]]
 ;
+; GFX9-LABEL: @sadd_sat_v2i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.sadd.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
+; GFX9-NEXT:    ret <2 x i16> [[TMP0]]
+;
 bb:
   %arg0.0 = extractelement <2 x i16> %arg0, i64 0
   %arg0.1 = extractelement <2 x i16> %arg0, i64 1
@@ -111,6 +126,11 @@ define <2 x i16> @ssub_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX8-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.ssub.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
 ; GFX8-NEXT:    ret <2 x i16> [[TMP0]]
 ;
+; GFX9-LABEL: @ssub_sat_v2i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.ssub.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
+; GFX9-NEXT:    ret <2 x i16> [[TMP0]]
+;
 bb:
   %arg0.0 = extractelement <2 x i16> %arg0, i64 0
   %arg0.1 = extractelement <2 x i16> %arg0, i64 1
@@ -252,6 +272,18 @@ define <3 x i16> @uadd_sat_v3i16(<3 x i16> %arg0, <3 x i16> %arg1) {
 ; GFX8-NEXT:    [[INS_2:%.*]] = insertelement <3 x i16> [[TMP3]], i16 [[ADD_2]], i64 2
 ; GFX8-NEXT:    ret <3 x i16> [[INS_2]]
 ;
+; GFX9-LABEL: @uadd_sat_v3i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[ARG0_2:%.*]] = extractelement <3 x i16> [[ARG0:%.*]], i64 2
+; GFX9-NEXT:    [[ARG1_2:%.*]] = extractelement <3 x i16> [[ARG1:%.*]], i64 2
+; GFX9-NEXT:    [[TMP0:%.*]] = shufflevector <3 x i16> [[ARG0]], <3 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:    [[TMP1:%.*]] = shufflevector <3 x i16> [[ARG1]], <3 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:    [[TMP2:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP0]], <2 x i16> [[TMP1]])
+; GFX9-NEXT:    [[ADD_2:%.*]] = call i16 @llvm.uadd.sat.i16(i16 [[ARG0_2]], i16 [[ARG1_2]])
+; GFX9-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> poison, <3 x i32> <i32 0, i32 1, i32 poison>
+; GFX9-NEXT:    [[INS_2:%.*]] = insertelement <3 x i16> [[TMP3]], i16 [[ADD_2]], i64 2
+; GFX9-NEXT:    ret <3 x i16> [[INS_2]]
+;
 bb:
   %arg0.0 = extractelement <3 x i16> %arg0, i64 0
   %arg0.1 = extractelement <3 x i16> %arg0, i64 1
@@ -291,19 +323,25 @@ define <4 x i16> @uadd_sat_v4i16(<4 x i16> %arg0, <4 x i16> %arg1) {
 ;
 ; GFX8-LABEL: @uadd_sat_v4i16(
 ; GFX8-NEXT:  bb:
-; GFX8-NEXT:    [[ARG0_2:%.*]] = extractelement <4 x i16> [[ARG0:%.*]], i64 2
-; GFX8-NEXT:    [[ARG0_3:%.*]] = extractelement <4 x i16> [[ARG0]], i64 3
-; GFX8-NEXT:    [[ARG1_2:%.*]] = extractelement <4 x i16> [[ARG1:%.*]], i64 2
-; GFX8-NEXT:    [[ARG1_3:%.*]] = extractelement <4 x i16> [[ARG1]], i64 3
-; GFX8-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i16> [[ARG0]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
-; GFX8-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i16> [[ARG1]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX8-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i16> [[ARG0:%.*]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX8-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i16> [[ARG1:%.*]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
 ; GFX8-NEXT:    [[TMP2:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP0]], <2 x i16> [[TMP1]])
-; GFX8-NEXT:    [[ADD_2:%.*]] = call i16 @llvm.uadd.sat.i16(i16 [[ARG0_2]], i16 [[ARG1_2]])
-; GFX8-NEXT:    [[ADD_3:%.*]] = call i16 @llvm.uadd.sat.i16(i16 [[ARG0_3]], i16 [[ARG1_3]])
-; GFX8-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; GFX8-NEXT:    [[INS_2:%.*]] = insertelement <4 x i16> [[TMP3]], i16 [[ADD_2]], i64 2
-; GFX8-NEXT:    [[INS_3:%.*]] = insertelement <4 x i16> [[INS_2]], i16 [[ADD_3]], i64 3
-; GFX8-NEXT:    ret <4 x i16> [[INS_3]]
+; GFX8-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i16> [[ARG0]], <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GFX8-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[ARG1]], <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GFX8-NEXT:    [[TMP5:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP3]], <2 x i16> [[TMP4]])
+; GFX8-NEXT:    [[INS_31:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX8-NEXT:    ret <4 x i16> [[INS_31]]
+;
+; GFX9-LABEL: @uadd_sat_v4i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i16> [[ARG0:%.*]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i16> [[ARG1:%.*]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:    [[TMP2:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP0]], <2 x i16> [[TMP1]])
+; GFX9-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i16> [[ARG0]], <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GFX9-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[ARG1]], <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GFX9-NEXT:    [[TMP5:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP3]], <2 x i16> [[TMP4]])
+; GFX9-NEXT:    [[INS_31:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:    ret <4 x i16> [[INS_31]]
 ;
 bb:
   %arg0.0 = extractelement <4 x i16> %arg0, i64 0
diff --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat.ll
index 2038400..0bb6413 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/add_sub_sat.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -passes=slp-vectorizer,instcombine %s | FileCheck -check-prefixes=GCN,GFX7 %s
 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -passes=slp-vectorizer,instcombine %s | FileCheck -check-prefixes=GCN,GFX8 %s
-; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -passes=slp-vectorizer,instcombine %s | FileCheck -check-prefixes=GCN,GFX8 %s
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -passes=slp-vectorizer,instcombine %s | FileCheck -check-prefixes=GCN,GFX9 %s
 
 define <2 x i16> @uadd_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX7-LABEL: @uadd_sat_v2i16(
@@ -21,6 +21,11 @@ define <2 x i16> @uadd_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX8-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
 ; GFX8-NEXT:    ret <2 x i16> [[TMP0]]
 ;
+; GFX9-LABEL: @uadd_sat_v2i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
+; GFX9-NEXT:    ret <2 x i16> [[TMP0]]
+;
 bb:
   %arg0.0 = extractelement <2 x i16> %arg0, i64 0
   %arg0.1 = extractelement <2 x i16> %arg0, i64 1
@@ -51,6 +56,11 @@ define <2 x i16> @usub_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX8-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.usub.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
 ; GFX8-NEXT:    ret <2 x i16> [[TMP0]]
 ;
+; GFX9-LABEL: @usub_sat_v2i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.usub.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
+; GFX9-NEXT:    ret <2 x i16> [[TMP0]]
+;
 bb:
   %arg0.0 = extractelement <2 x i16> %arg0, i64 0
   %arg0.1 = extractelement <2 x i16> %arg0, i64 1
@@ -81,6 +91,11 @@ define <2 x i16> @sadd_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX8-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.sadd.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
 ; GFX8-NEXT:    ret <2 x i16> [[TMP0]]
 ;
+; GFX9-LABEL: @sadd_sat_v2i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.sadd.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
+; GFX9-NEXT:    ret <2 x i16> [[TMP0]]
+;
 bb:
   %arg0.0 = extractelement <2 x i16> %arg0, i64 0
   %arg0.1 = extractelement <2 x i16> %arg0, i64 1
@@ -111,6 +126,11 @@ define <2 x i16> @ssub_sat_v2i16(<2 x i16> %arg0, <2 x i16> %arg1) {
 ; GFX8-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.ssub.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
 ; GFX8-NEXT:    ret <2 x i16> [[TMP0]]
 ;
+; GFX9-LABEL: @ssub_sat_v2i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = call <2 x i16> @llvm.ssub.sat.v2i16(<2 x i16> [[ARG0:%.*]], <2 x i16> [[ARG1:%.*]])
+; GFX9-NEXT:    ret <2 x i16> [[TMP0]]
+;
 bb:
   %arg0.0 = extractelement <2 x i16> %arg0, i64 0
   %arg0.1 = extractelement <2 x i16> %arg0, i64 1
@@ -252,6 +272,18 @@ define <3 x i16> @uadd_sat_v3i16(<3 x i16> %arg0, <3 x i16> %arg1) {
 ; GFX8-NEXT:    [[INS_2:%.*]] = insertelement <3 x i16> [[TMP3]], i16 [[ADD_2]], i64 2
 ; GFX8-NEXT:    ret <3 x i16> [[INS_2]]
 ;
+; GFX9-LABEL: @uadd_sat_v3i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[ARG0_2:%.*]] = extractelement <3 x i16> [[ARG0:%.*]], i64 2
+; GFX9-NEXT:    [[ARG1_2:%.*]] = extractelement <3 x i16> [[ARG1:%.*]], i64 2
+; GFX9-NEXT:    [[TMP0:%.*]] = shufflevector <3 x i16> [[ARG0]], <3 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:    [[TMP1:%.*]] = shufflevector <3 x i16> [[ARG1]], <3 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:    [[TMP2:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP0]], <2 x i16> [[TMP1]])
+; GFX9-NEXT:    [[ADD_2:%.*]] = call i16 @llvm.uadd.sat.i16(i16 [[ARG0_2]], i16 [[ARG1_2]])
+; GFX9-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> poison, <3 x i32> <i32 0, i32 1, i32 poison>
+; GFX9-NEXT:    [[INS_2:%.*]] = insertelement <3 x i16> [[TMP3]], i16 [[ADD_2]], i64 2
+; GFX9-NEXT:    ret <3 x i16> [[INS_2]]
+;
 bb:
   %arg0.0 = extractelement <3 x i16> %arg0, i64 0
   %arg0.1 = extractelement <3 x i16> %arg0, i64 1
@@ -291,19 +323,25 @@ define <4 x i16> @uadd_sat_v4i16(<4 x i16> %arg0, <4 x i16> %arg1) {
 ;
 ; GFX8-LABEL: @uadd_sat_v4i16(
 ; GFX8-NEXT:  bb:
-; GFX8-NEXT:    [[ARG0_2:%.*]] = extractelement <4 x i16> [[ARG0:%.*]], i64 2
-; GFX8-NEXT:    [[ARG0_3:%.*]] = extractelement <4 x i16> [[ARG0]], i64 3
-; GFX8-NEXT:    [[ARG1_2:%.*]] = extractelement <4 x i16> [[ARG1:%.*]], i64 2
-; GFX8-NEXT:    [[ARG1_3:%.*]] = extractelement <4 x i16> [[ARG1]], i64 3
-; GFX8-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i16> [[ARG0]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
-; GFX8-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i16> [[ARG1]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX8-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i16> [[ARG0:%.*]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX8-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i16> [[ARG1:%.*]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
 ; GFX8-NEXT:    [[TMP2:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP0]], <2 x i16> [[TMP1]])
-; GFX8-NEXT:    [[ADD_2:%.*]] = call i16 @llvm.uadd.sat.i16(i16 [[ARG0_2]], i16 [[ARG1_2]])
-; GFX8-NEXT:    [[ADD_3:%.*]] = call i16 @llvm.uadd.sat.i16(i16 [[ARG0_3]], i16 [[ARG1_3]])
-; GFX8-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; GFX8-NEXT:    [[INS_2:%.*]] = insertelement <4 x i16> [[TMP3]], i16 [[ADD_2]], i64 2
-; GFX8-NEXT:    [[INS_3:%.*]] = insertelement <4 x i16> [[INS_2]], i16 [[ADD_3]], i64 3
-; GFX8-NEXT:    ret <4 x i16> [[INS_3]]
+; GFX8-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i16> [[ARG0]], <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GFX8-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[ARG1]], <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GFX8-NEXT:    [[TMP5:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP3]], <2 x i16> [[TMP4]])
+; GFX8-NEXT:    [[INS_31:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX8-NEXT:    ret <4 x i16> [[INS_31]]
+;
+; GFX9-LABEL: @uadd_sat_v4i16(
+; GFX9-NEXT:  bb:
+; GFX9-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i16> [[ARG0:%.*]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i16> [[ARG1:%.*]], <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GFX9-NEXT:    [[TMP2:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP0]], <2 x i16> [[TMP1]])
+; GFX9-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i16> [[ARG0]], <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GFX9-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[ARG1]], <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GFX9-NEXT:    [[TMP5:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP3]], <2 x i16> [[TMP4]])
+; GFX9-NEXT:    [[INS_31:%.*]] = shufflevector <2 x i16> [[TMP2]], <2 x i16> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GFX9-NEXT:    ret <4 x i16> [[INS_31]]
 ;
 bb:
   %arg0.0 = extractelement <4 x i16> %arg0, i64 0
diff --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/crash_extract_subvector_cost.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/crash_extract_subvector_cost.ll
index 0a020c8..e474bab 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/crash_extract_subvector_cost.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/crash_extract_subvector_cost.ll
@@ -4,15 +4,10 @@
 define <2 x i16> @uadd_sat_v9i16_combine_vi16(<9 x i16> %arg0, <9 x i16> %arg1) {
 ; CHECK-LABEL: @uadd_sat_v9i16_combine_vi16(
 ; CHECK-NEXT:  bb:
-; CHECK-NEXT:    [[ARG0_1:%.*]] = extractelement <9 x i16> undef, i64 7
-; CHECK-NEXT:    [[ARG0_2:%.*]] = extractelement <9 x i16> [[ARG0:%.*]], i64 8
-; CHECK-NEXT:    [[ARG1_1:%.*]] = extractelement <9 x i16> [[ARG1:%.*]], i64 7
-; CHECK-NEXT:    [[ARG1_2:%.*]] = extractelement <9 x i16> [[ARG1]], i64 8
-; CHECK-NEXT:    [[ADD_1:%.*]] = call i16 @llvm.uadd.sat.i16(i16 [[ARG0_1]], i16 [[ARG1_1]])
-; CHECK-NEXT:    [[ADD_2:%.*]] = call i16 @llvm.uadd.sat.i16(i16 [[ARG0_2]], i16 [[ARG1_2]])
-; CHECK-NEXT:    [[INS_1:%.*]] = insertelement <2 x i16> undef, i16 [[ADD_1]], i64 0
-; CHECK-NEXT:    [[INS_2:%.*]] = insertelement <2 x i16> [[INS_1]], i16 [[ADD_2]], i64 1
-; CHECK-NEXT:    ret <2 x i16> [[INS_2]]
+; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <9 x i16> [[ARG0:%.*]], <9 x i16> poison, <2 x i32> <i32 poison, i32 8>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <9 x i16> [[ARG1:%.*]], <9 x i16> poison, <2 x i32> <i32 7, i32 8>
+; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> [[TMP0]], <2 x i16> [[TMP1]])
+; CHECK-NEXT:    ret <2 x i16> [[TMP2]]
 ;
 bb:
   %arg0.1 = extractelement <9 x i16> undef, i64 7
diff --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/phi-result-use-order.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/phi-result-use-order.ll
index 46980b33..3b63c1e 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/phi-result-use-order.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/phi-result-use-order.ll
@@ -4,23 +4,20 @@
 define <4 x half> @phis(i1 %cmp1, <4 x half> %in1, <4 x half> %in2)  {
 ; CHECK-LABEL: @phis(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[A2:%.*]] = extractelement <4 x half> [[IN1:%.*]], i64 2
-; CHECK-NEXT:    [[A3:%.*]] = extractelement <4 x half> [[IN1]], i64 3
-; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x half> [[IN1]], <4 x half> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x half> [[IN1:%.*]], <4 x half> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x half> [[IN1]], <4 x half> poison, <2 x i32> <i32 2, i32 3>
 ; CHECK-NEXT:    br i1 [[CMP1:%.*]], label [[BB1:%.*]], label [[BB0:%.*]]
 ; CHECK:       bb0:
-; CHECK-NEXT:    [[B2:%.*]] = extractelement <4 x half> [[IN2:%.*]], i64 2
-; CHECK-NEXT:    [[B3:%.*]] = extractelement <4 x half> [[IN2]], i64 3
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x half> [[IN2]], <4 x half> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x half> [[IN2:%.*]], <4 x half> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x half> [[IN2]], <4 x half> poison, <2 x i32> <i32 2, i32 3>
 ; CHECK-NEXT:    br label [[BB1]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[C2:%.*]] = phi half [ [[A2]], [[ENTRY:%.*]] ], [ [[B2]], [[BB0]] ]
-; CHECK-NEXT:    [[C3:%.*]] = phi half [ [[A3]], [[ENTRY]] ], [ [[B3]], [[BB0]] ]
-; CHECK-NEXT:    [[TMP2:%.*]] = phi <2 x half> [ [[TMP0]], [[ENTRY]] ], [ [[TMP1]], [[BB0]] ]
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x half> [[TMP2]], <2 x half> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[O2:%.*]] = insertelement <4 x half> [[TMP3]], half [[C2]], i64 2
-; CHECK-NEXT:    [[O3:%.*]] = insertelement <4 x half> [[O2]], half [[C3]], i64 3
-; CHECK-NEXT:    ret <4 x half> [[O3]]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <2 x half> [ [[TMP0]], [[ENTRY:%.*]] ], [ [[TMP2]], [[BB0]] ]
+; CHECK-NEXT:    [[TMP5:%.*]] = phi <2 x half> [ [[TMP1]], [[ENTRY]] ], [ [[TMP3]], [[BB0]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x half> [[TMP4]], <2 x half> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <2 x half> [[TMP5]], <2 x half> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <2 x half> [[TMP4]], <2 x half> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    ret <4 x half> [[TMP8]]
 ;
 entry:
   %a0 = extractelement <4 x half> %in1, i64 0
@@ -52,23 +49,20 @@ bb1:
 define <4 x half> @phis_reverse(i1 %cmp1, <4 x half> %in1, <4 x half> %in2)  {
 ; CHECK-LABEL: @phis_reverse(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[A2:%.*]] = extractelement <4 x half> [[IN1:%.*]], i64 2
-; CHECK-NEXT:    [[A3:%.*]] = extractelement <4 x half> [[IN1]], i64 3
-; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x half> [[IN1]], <4 x half> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x half> [[IN1:%.*]], <4 x half> poison, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x half> [[IN1]], <4 x half> poison, <2 x i32> <i32 0, i32 1>
 ; CHECK-NEXT:    br i1 [[CMP1:%.*]], label [[BB1:%.*]], label [[BB0:%.*]]
 ; CHECK:       bb0:
-; CHECK-NEXT:    [[B2:%.*]] = extractelement <4 x half> [[IN2:%.*]], i64 2
-; CHECK-NEXT:    [[B3:%.*]] = extractelement <4 x half> [[IN2]], i64 3
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x half> [[IN2]], <4 x half> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x half> [[IN2:%.*]], <4 x half> poison, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x half> [[IN2]], <4 x half> poison, <2 x i32> <i32 0, i32 1>
 ; CHECK-NEXT:    br label [[BB1]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[C3:%.*]] = phi half [ [[A3]], [[ENTRY:%.*]] ], [ [[B3]], [[BB0]] ]
-; CHECK-NEXT:    [[C2:%.*]] = phi half [ [[A2]], [[ENTRY]] ], [ [[B2]], [[BB0]] ]
-; CHECK-NEXT:    [[TMP2:%.*]] = phi <2 x half> [ [[TMP0]], [[ENTRY]] ], [ [[TMP1]], [[BB0]] ]
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x half> [[TMP2]], <2 x half> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[O2:%.*]] = insertelement <4 x half> [[TMP3]], half [[C2]], i64 2
-; CHECK-NEXT:    [[O3:%.*]] = insertelement <4 x half> [[O2]], half [[C3]], i64 3
-; CHECK-NEXT:    ret <4 x half> [[O3]]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <2 x half> [ [[TMP0]], [[ENTRY:%.*]] ], [ [[TMP2]], [[BB0]] ]
+; CHECK-NEXT:    [[TMP5:%.*]] = phi <2 x half> [ [[TMP1]], [[ENTRY]] ], [ [[TMP3]], [[BB0]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x half> [[TMP5]], <2 x half> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <2 x half> [[TMP4]], <2 x half> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <4 x half> [[TMP6]], <4 x half> [[TMP7]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    ret <4 x half> [[TMP8]]
 ;
 entry:
   %a0 = extractelement <4 x half> %in1, i64 0
diff --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/reduction.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/reduction.ll
index b34b9a3..dfa8be9 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/reduction.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/reduction.ll
@@ -3,21 +3,10 @@
 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -passes=slp-vectorizer,dce < %s | FileCheck -check-prefixes=GCN,VI %s
 
 define half @reduction_half4(<4 x half> %a) {
-; GFX9-LABEL: @reduction_half4(
-; GFX9-NEXT:  entry:
-; GFX9-NEXT:    [[TMP0:%.*]] = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH8000, <4 x half> [[A:%.*]])
-; GFX9-NEXT:    ret half [[TMP0]]
-;
-; VI-LABEL: @reduction_half4(
-; VI-NEXT:  entry:
-; VI-NEXT:    [[ELT0:%.*]] = extractelement <4 x half> [[A:%.*]], i64 0
-; VI-NEXT:    [[ELT1:%.*]] = extractelement <4 x half> [[A]], i64 1
-; VI-NEXT:    [[ELT2:%.*]] = extractelement <4 x half> [[A]], i64 2
-; VI-NEXT:    [[ELT3:%.*]] = extractelement <4 x half> [[A]], i64 3
-; VI-NEXT:    [[ADD1:%.*]] = fadd fast half [[ELT1]], [[ELT0]]
-; VI-NEXT:    [[ADD2:%.*]] = fadd fast half [[ELT2]], [[ADD1]]
-; VI-NEXT:    [[ADD3:%.*]] = fadd fast half [[ELT3]], [[ADD2]]
-; VI-NEXT:    ret half [[ADD3]]
+; GCN-LABEL: @reduction_half4(
+; GCN-NEXT:  entry:
+; GCN-NEXT:    [[TMP0:%.*]] = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH8000, <4 x half> [[A:%.*]])
+; GCN-NEXT:    ret half [[TMP0]]
 ;
 entry:
   %elt0 = extractelement <4 x half> %a, i64 0
@@ -33,29 +22,10 @@ entry:
 }
 
 define half @reduction_half8(<8 x half> %vec8) {
-; GFX9-LABEL: @reduction_half8(
-; GFX9-NEXT:  entry:
-; GFX9-NEXT:    [[TMP0:%.*]] = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH8000, <8 x half> [[VEC8:%.*]])
-; GFX9-NEXT:    ret half [[TMP0]]
-;
-; VI-LABEL: @reduction_half8(
-; VI-NEXT:  entry:
-; VI-NEXT:    [[ELT0:%.*]] = extractelement <8 x half> [[VEC8:%.*]], i64 0
-; VI-NEXT:    [[ELT1:%.*]] = extractelement <8 x half> [[VEC8]], i64 1
-; VI-NEXT:    [[ELT2:%.*]] = extractelement <8 x half> [[VEC8]], i64 2
-; VI-NEXT:    [[ELT3:%.*]] = extractelement <8 x half> [[VEC8]], i64 3
-; VI-NEXT:    [[ELT4:%.*]] = extractelement <8 x half> [[VEC8]], i64 4
-; VI-NEXT:    [[ELT5:%.*]] = extractelement <8 x half> [[VEC8]], i64 5
-; VI-NEXT:    [[ELT6:%.*]] = extractelement <8 x half> [[VEC8]], i64 6
-; VI-NEXT:    [[ELT7:%.*]] = extractelement <8 x half> [[VEC8]], i64 7
-; VI-NEXT:    [[ADD1:%.*]] = fadd fast half [[ELT1]], [[ELT0]]
-; VI-NEXT:    [[ADD2:%.*]] = fadd fast half [[ELT2]], [[ADD1]]
-; VI-NEXT:    [[ADD3:%.*]] = fadd fast half [[ELT3]], [[ADD2]]
-; VI-NEXT:    [[ADD4:%.*]] = fadd fast half [[ELT4]], [[ADD3]]
-; VI-NEXT:    [[ADD5:%.*]] = fadd fast half [[ELT5]], [[ADD4]]
-; VI-NEXT:    [[ADD6:%.*]] = fadd fast half [[ELT6]], [[ADD5]]
-; VI-NEXT:    [[ADD7:%.*]] = fadd fast half [[ELT7]], [[ADD6]]
-; VI-NEXT:    ret half [[ADD7]]
+; GCN-LABEL: @reduction_half8(
+; GCN-NEXT:  entry:
+; GCN-NEXT:    [[TMP0:%.*]] = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH8000, <8 x half> [[VEC8:%.*]])
+; GCN-NEXT:    ret half [[TMP0]]
 ;
 entry:
   %elt0 = extractelement <8 x half> %vec8, i64 0
@@ -86,15 +56,7 @@ define half @reduction_half16(<16 x half> %vec16) {
 ;
 ; VI-LABEL: @reduction_half16(
 ; VI-NEXT:  entry:
-; VI-NEXT:    [[ELT0:%.*]] = extractelement <16 x half> [[VEC16:%.*]], i64 0
-; VI-NEXT:    [[ELT1:%.*]] = extractelement <16 x half> [[VEC16]], i64 1
-; VI-NEXT:    [[ELT2:%.*]] = extractelement <16 x half> [[VEC16]], i64 2
-; VI-NEXT:    [[ELT3:%.*]] = extractelement <16 x half> [[VEC16]], i64 3
-; VI-NEXT:    [[ELT4:%.*]] = extractelement <16 x half> [[VEC16]], i64 4
-; VI-NEXT:    [[ELT5:%.*]] = extractelement <16 x half> [[VEC16]], i64 5
-; VI-NEXT:    [[ELT6:%.*]] = extractelement <16 x half> [[VEC16]], i64 6
-; VI-NEXT:    [[ELT7:%.*]] = extractelement <16 x half> [[VEC16]], i64 7
-; VI-NEXT:    [[ELT8:%.*]] = extractelement <16 x half> [[VEC16]], i64 8
+; VI-NEXT:    [[ELT8:%.*]] = extractelement <16 x half> [[VEC16:%.*]], i64 8
 ; VI-NEXT:    [[ELT9:%.*]] = extractelement <16 x half> [[VEC16]], i64 9
 ; VI-NEXT:    [[ELT10:%.*]] = extractelement <16 x half> [[VEC16]], i64 10
 ; VI-NEXT:    [[ELT11:%.*]] = extractelement <16 x half> [[VEC16]], i64 11
@@ -102,22 +64,17 @@ define half @reduction_half16(<16 x half> %vec16) {
 ; VI-NEXT:    [[ELT13:%.*]] = extractelement <16 x half> [[VEC16]], i64 13
 ; VI-NEXT:    [[ELT14:%.*]] = extractelement <16 x half> [[VEC16]], i64 14
 ; VI-NEXT:    [[ELT15:%.*]] = extractelement <16 x half> [[VEC16]], i64 15
-; VI-NEXT:    [[ADD1:%.*]] = fadd fast half [[ELT1]], [[ELT0]]
-; VI-NEXT:    [[ADD2:%.*]] = fadd fast half [[ELT2]], [[ADD1]]
-; VI-NEXT:    [[ADD3:%.*]] = fadd fast half [[ELT3]], [[ADD2]]
-; VI-NEXT:    [[ADD4:%.*]] = fadd fast half [[ELT4]], [[ADD3]]
-; VI-NEXT:    [[ADD5:%.*]] = fadd fast half [[ELT5]], [[ADD4]]
-; VI-NEXT:    [[ADD6:%.*]] = fadd fast half [[ELT6]], [[ADD5]]
-; VI-NEXT:    [[ADD7:%.*]] = fadd fast half [[ELT7]], [[ADD6]]
-; VI-NEXT:    [[ADD8:%.*]] = fadd fast half [[ELT8]], [[ADD7]]
-; VI-NEXT:    [[ADD9:%.*]] = fadd fast half [[ELT9]], [[ADD8]]
-; VI-NEXT:    [[ADD10:%.*]] = fadd fast half [[ELT10]], [[ADD9]]
-; VI-NEXT:    [[ADD11:%.*]] = fadd fast half [[ELT11]], [[ADD10]]
-; VI-NEXT:    [[ADD12:%.*]] = fadd fast half [[ELT12]], [[ADD11]]
-; VI-NEXT:    [[ADD13:%.*]] = fadd fast half [[ELT13]], [[ADD12]]
-; VI-NEXT:    [[ADD14:%.*]] = fadd fast half [[ELT14]], [[ADD13]]
-; VI-NEXT:    [[ADD15:%.*]] = fadd fast half [[ELT15]], [[ADD14]]
-; VI-NEXT:    ret half [[ADD15]]
+; VI-NEXT:    [[TMP0:%.*]] = shufflevector <16 x half> [[VEC16]], <16 x half> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VI-NEXT:    [[TMP1:%.*]] = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH8000, <8 x half> [[TMP0]])
+; VI-NEXT:    [[OP_RDX:%.*]] = fadd fast half [[TMP1]], [[ELT8]]
+; VI-NEXT:    [[OP_RDX1:%.*]] = fadd fast half [[ELT9]], [[ELT10]]
+; VI-NEXT:    [[OP_RDX2:%.*]] = fadd fast half [[ELT11]], [[ELT12]]
+; VI-NEXT:    [[OP_RDX3:%.*]] = fadd fast half [[ELT13]], [[ELT14]]
+; VI-NEXT:    [[OP_RDX4:%.*]] = fadd fast half [[OP_RDX]], [[OP_RDX1]]
+; VI-NEXT:    [[OP_RDX5:%.*]] = fadd fast half [[OP_RDX2]], [[OP_RDX3]]
+; VI-NEXT:    [[OP_RDX6:%.*]] = fadd fast half [[OP_RDX4]], [[OP_RDX5]]
+; VI-NEXT:    [[OP_RDX7:%.*]] = fadd fast half [[OP_RDX6]], [[ELT15]]
+; VI-NEXT:    ret half [[OP_RDX7]]
 ;
 entry:
   %elt0 = extractelement <16 x half> %vec16, i64 0
@@ -183,21 +140,10 @@ entry:
 }
 
 define i16 @reduction_v4i16(<4 x i16> %a) {
-; GFX9-LABEL: @reduction_v4i16(
-; GFX9-NEXT:  entry:
-; GFX9-NEXT:    [[TMP0:%.*]] = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> [[A:%.*]])
-; GFX9-NEXT:    ret i16 [[TMP0]]
-;
-; VI-LABEL: @reduction_v4i16(
-; VI-NEXT:  entry:
-; VI-NEXT:    [[ELT0:%.*]] = extractelement <4 x i16> [[A:%.*]], i64 0
-; VI-NEXT:    [[ELT1:%.*]] = extractelement <4 x i16> [[A]], i64 1
-; VI-NEXT:    [[ELT2:%.*]] = extractelement <4 x i16> [[A]], i64 2
-; VI-NEXT:    [[ELT3:%.*]] = extractelement <4 x i16> [[A]], i64 3
-; VI-NEXT:    [[ADD1:%.*]] = add i16 [[ELT1]], [[ELT0]]
-; VI-NEXT:    [[ADD2:%.*]] = add i16 [[ELT2]], [[ADD1]]
-; VI-NEXT:    [[ADD3:%.*]] = add i16 [[ELT3]], [[ADD2]]
-; VI-NEXT:    ret i16 [[ADD3]]
+; GCN-LABEL: @reduction_v4i16(
+; GCN-NEXT:  entry:
+; GCN-NEXT:    [[TMP0:%.*]] = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> [[A:%.*]])
+; GCN-NEXT:    ret i16 [[TMP0]]
 ;
 entry:
   %elt0 = extractelement <4 x i16> %a, i64 0
@@ -213,29 +159,10 @@ entry:
 }
 
 define i16 @reduction_v8i16(<8 x i16> %vec8) {
-; GFX9-LABEL: @reduction_v8i16(
-; GFX9-NEXT:  entry:
-; GFX9-NEXT:    [[TMP0:%.*]] = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> [[VEC8:%.*]])
-; GFX9-NEXT:    ret i16 [[TMP0]]
-;
-; VI-LABEL: @reduction_v8i16(
-; VI-NEXT:  entry:
-; VI-NEXT:    [[ELT0:%.*]] = extractelement <8 x i16> [[VEC8:%.*]], i64 0
-; VI-NEXT:    [[ELT1:%.*]] = extractelement <8 x i16> [[VEC8]], i64 1
-; VI-NEXT:    [[ELT2:%.*]] = extractelement <8 x i16> [[VEC8]], i64 2
-; VI-NEXT:    [[ELT3:%.*]] = extractelement <8 x i16> [[VEC8]], i64 3
-; VI-NEXT:    [[ELT4:%.*]] = extractelement <8 x i16> [[VEC8]], i64 4
-; VI-NEXT:    [[ELT5:%.*]] = extractelement <8 x i16> [[VEC8]], i64 5
-; VI-NEXT:    [[ELT6:%.*]] = extractelement <8 x i16> [[VEC8]], i64 6
-; VI-NEXT:    [[ELT7:%.*]] = extractelement <8 x i16> [[VEC8]], i64 7
-; VI-NEXT:    [[ADD1:%.*]] = add i16 [[ELT1]], [[ELT0]]
-; VI-NEXT:    [[ADD2:%.*]] = add i16 [[ELT2]], [[ADD1]]
-; VI-NEXT:    [[ADD3:%.*]] = add i16 [[ELT3]], [[ADD2]]
-; VI-NEXT:    [[ADD4:%.*]] = add i16 [[ELT4]], [[ADD3]]
-; VI-NEXT:    [[ADD5:%.*]] = add i16 [[ELT5]], [[ADD4]]
-; VI-NEXT:    [[ADD6:%.*]] = add i16 [[ELT6]], [[ADD5]]
-; VI-NEXT:    [[ADD7:%.*]] = add i16 [[ELT7]], [[ADD6]]
-; VI-NEXT:    ret i16 [[ADD7]]
+; GCN-LABEL: @reduction_v8i16(
+; GCN-NEXT:  entry:
+; GCN-NEXT:    [[TMP0:%.*]] = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> [[VEC8:%.*]])
+; GCN-NEXT:    ret i16 [[TMP0]]
 ;
 entry:
   %elt0 = extractelement <8 x i16> %vec8, i64 0