diff options
Diffstat (limited to 'llvm/test/Transforms')
21 files changed, 1266 insertions, 81 deletions
diff --git a/llvm/test/Transforms/Coroutines/coro-catchswitch-cleanuppad.ll b/llvm/test/Transforms/Coroutines/coro-catchswitch-cleanuppad.ll index d0e7c1c2..e1e1611 100644 --- a/llvm/test/Transforms/Coroutines/coro-catchswitch-cleanuppad.ll +++ b/llvm/test/Transforms/Coroutines/coro-catchswitch-cleanuppad.ll @@ -80,8 +80,8 @@ cleanup2: ; CHECK: cleanup2.corodispatch: ; CHECK: %1 = phi i8 [ 0, %handler2 ], [ 1, %catch.dispatch.2 ] ; CHECK: %2 = cleanuppad within %h1 [] -; CHECK: %switch = icmp ult i8 %1, 1 -; CHECK: br i1 %switch, label %cleanup2.from.handler2, label %cleanup2.from.catch.dispatch.2 +; CHECK: %3 = icmp eq i8 %1, 0 +; CHECK: br i1 %3, label %cleanup2.from.handler2, label %cleanup2.from.catch.dispatch.2 ; CHECK: cleanup2.from.handler2: ; CHECK: %valueB.reload = load i32, ptr %valueB.spill.addr, align 4 diff --git a/llvm/test/Transforms/FunctionAttrs/norecurse_libfunc_address_taken.ll b/llvm/test/Transforms/FunctionAttrs/norecurse_libfunc_address_taken.ll new file mode 100644 index 0000000..bcdf75b --- /dev/null +++ b/llvm/test/Transforms/FunctionAttrs/norecurse_libfunc_address_taken.ll @@ -0,0 +1,40 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 5 +; RUN: opt < %s -passes=norecurse-lto-inference -S | FileCheck %s + +; This test includes a call to a library function which is not marked as +; NoCallback. Function bob() does not have internal linkage and hence prevents +; norecurse to be added. + +@.str = private unnamed_addr constant [12 x i8] c"Hello World\00", align 1 + +;. +; CHECK: @.str = private unnamed_addr constant [12 x i8] c"Hello World\00", align 1 +;. +define dso_local void @bob() { +; CHECK-LABEL: define dso_local void @bob() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: [[CALL:%.*]] = tail call i32 (ptr, ...) @printf(ptr nonnull dereferenceable(1) @.str) +; CHECK-NEXT: ret void +; +entry: + %call = tail call i32 (ptr, ...) @printf(ptr nonnull dereferenceable(1) @.str) + ret void +} + +declare i32 @printf(ptr readonly captures(none), ...) + +define dso_local i32 @main() norecurse { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define dso_local i32 @main( +; CHECK-SAME: ) #[[ATTR0:[0-9]+]] { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bob() +; CHECK-NEXT: ret i32 0 +; +entry: + tail call void @bob() + ret i32 0 +} +;. +; CHECK: attributes #[[ATTR0]] = { norecurse } +;. diff --git a/llvm/test/Transforms/FunctionAttrs/norecurse_libfunc_no_address_taken.ll b/llvm/test/Transforms/FunctionAttrs/norecurse_libfunc_no_address_taken.ll new file mode 100644 index 0000000..a03b4ca --- /dev/null +++ b/llvm/test/Transforms/FunctionAttrs/norecurse_libfunc_no_address_taken.ll @@ -0,0 +1,45 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 5 +; RUN: opt < %s -passes=norecurse-lto-inference -S | FileCheck %s + +; This test includes a call to a library function which is not marked as +; NoCallback. All functions except main() are internal and main is marked +; norecurse, so as to not block norecurse to be added to bob(). + +@.str = private unnamed_addr constant [12 x i8] c"Hello World\00", align 1 + +; Function Attrs: nofree noinline nounwind uwtable +;. +; CHECK: @.str = private unnamed_addr constant [12 x i8] c"Hello World\00", align 1 +;. +define internal void @bob() { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define internal void @bob( +; CHECK-SAME: ) #[[ATTR0:[0-9]+]] { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: [[CALL:%.*]] = tail call i32 (ptr, ...) @printf(ptr nonnull dereferenceable(1) @.str) +; CHECK-NEXT: ret void +; +entry: + %call = tail call i32 (ptr, ...) @printf(ptr nonnull dereferenceable(1) @.str) + ret void +} + +; Function Attrs: nofree nounwind +declare i32 @printf(ptr readonly captures(none), ...) + +; Function Attrs: nofree norecurse nounwind uwtable +define dso_local i32 @main() norecurse { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define dso_local i32 @main( +; CHECK-SAME: ) #[[ATTR0]] { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bob() +; CHECK-NEXT: ret i32 0 +; +entry: + tail call void @bob() + ret i32 0 +} +;. +; CHECK: attributes #[[ATTR0]] = { norecurse } +;. diff --git a/llvm/test/Transforms/FunctionAttrs/norecurse_lto.ll b/llvm/test/Transforms/FunctionAttrs/norecurse_lto.ll new file mode 100644 index 0000000..5be707b --- /dev/null +++ b/llvm/test/Transforms/FunctionAttrs/norecurse_lto.ll @@ -0,0 +1,69 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 5 +; RUN: opt < %s -passes=norecurse-lto-inference -S | FileCheck %s + +; This test includes a call graph which has a recursive function(foo2) which +; calls a non-recursive internal function (foo3) satisfying the norecurse +; attribute criteria. + + +define internal void @foo3() { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define internal void @foo3( +; CHECK-SAME: ) #[[ATTR0:[0-9]+]] { +; CHECK-NEXT: ret void +; + ret void +} + +define internal i32 @foo2(i32 %accum, i32 %n) { +; CHECK-LABEL: define internal i32 @foo2( +; CHECK-SAME: i32 [[ACCUM:%.*]], i32 [[N:%.*]]) { +; CHECK-NEXT: [[ENTRY:.*]]: +; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[N]], 0 +; CHECK-NEXT: br i1 [[CMP]], label %[[EXIT:.*]], label %[[RECURSE:.*]] +; CHECK: [[RECURSE]]: +; CHECK-NEXT: [[SUB:%.*]] = sub i32 [[N]], 1 +; CHECK-NEXT: [[MUL:%.*]] = mul i32 [[ACCUM]], [[SUB]] +; CHECK-NEXT: [[CALL:%.*]] = call i32 @foo2(i32 [[MUL]], i32 [[SUB]]) +; CHECK-NEXT: call void @foo3() +; CHECK-NEXT: br label %[[EXIT]] +; CHECK: [[EXIT]]: +; CHECK-NEXT: [[RES:%.*]] = phi i32 [ [[ACCUM]], %[[ENTRY]] ], [ [[CALL]], %[[RECURSE]] ] +; CHECK-NEXT: ret i32 [[RES]] +; +entry: + %cmp = icmp eq i32 %n, 0 + br i1 %cmp, label %exit, label %recurse + +recurse: + %sub = sub i32 %n, 1 + %mul = mul i32 %accum, %sub + %call = call i32 @foo2(i32 %mul, i32 %sub) + call void @foo3() + br label %exit + +exit: + %res = phi i32 [ %accum, %entry ], [ %call, %recurse ] + ret i32 %res +} + +define internal i32 @foo1() { +; CHECK-LABEL: define internal i32 @foo1() { +; CHECK-NEXT: [[RES:%.*]] = call i32 @foo2(i32 1, i32 5) +; CHECK-NEXT: ret i32 [[RES]] +; + %res = call i32 @foo2(i32 1, i32 5) + ret i32 %res +} + +define dso_local i32 @main() { +; CHECK-LABEL: define dso_local i32 @main() { +; CHECK-NEXT: [[RES:%.*]] = call i32 @foo1() +; CHECK-NEXT: ret i32 [[RES]] +; + %res = call i32 @foo1() + ret i32 %res +} +;. +; CHECK: attributes #[[ATTR0]] = { norecurse } +;. diff --git a/llvm/test/Transforms/FunctionAttrs/norecurse_multi_scc_indirect_recursion.ll b/llvm/test/Transforms/FunctionAttrs/norecurse_multi_scc_indirect_recursion.ll new file mode 100644 index 0000000..e351f60 --- /dev/null +++ b/llvm/test/Transforms/FunctionAttrs/norecurse_multi_scc_indirect_recursion.ll @@ -0,0 +1,141 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 5 +; RUN: opt < %s -passes=norecurse-lto-inference -S | FileCheck %s + +; This test includes a call graph with multiple SCCs. The purpose of this is +; to check that norecurse is not added when a function is part of non-singular +; SCC. +; There are three different SCCs in this test: +; SCC#1: f1, foo, bar, foo1, bar1 +; SCC#2: bar2, bar3, bar4 +; SCC#3: baz, fun +; None of these functions should be marked as norecurse + +define internal void @bar1() { +; CHECK-LABEL: define internal void @bar1() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @f1() +; CHECK-NEXT: ret void +; +entry: + tail call void @f1() + ret void +} + +define internal void @f1() { +; CHECK-LABEL: define internal void @f1() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @foo() +; CHECK-NEXT: tail call void @bar2() +; CHECK-NEXT: tail call void @baz() +; CHECK-NEXT: ret void +; +entry: + tail call void @foo() + tail call void @bar2() + tail call void @baz() + ret void +} + +define dso_local i32 @main() norecurse { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define dso_local i32 @main( +; CHECK-SAME: ) #[[ATTR0:[0-9]+]] { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @f1() +; CHECK-NEXT: ret i32 0 +; +entry: + tail call void @f1() + ret i32 0 +} + +define internal void @foo1() { +; CHECK-LABEL: define internal void @foo1() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar1() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar1() + ret void +} + +define internal void @bar() { +; CHECK-LABEL: define internal void @bar() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @foo1() +; CHECK-NEXT: ret void +; +entry: + tail call void @foo1() + ret void +} + +define internal void @foo() { +; CHECK-LABEL: define internal void @foo() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar() + ret void +} + +define internal void @bar4() { +; CHECK-LABEL: define internal void @bar4() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar2() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar2() + ret void +} + +define internal void @bar2() { +; CHECK-LABEL: define internal void @bar2() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar3() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar3() + ret void +} + +define internal void @bar3() { +; CHECK-LABEL: define internal void @bar3() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar4() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar4() + ret void +} + +define internal void @fun() { +; CHECK-LABEL: define internal void @fun() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @baz() +; CHECK-NEXT: ret void +; +entry: + tail call void @baz() + ret void +} + +define internal void @baz() { +; CHECK-LABEL: define internal void @baz() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @fun() +; CHECK-NEXT: ret void +; +entry: + tail call void @fun() + ret void +} +;. +; CHECK: attributes #[[ATTR0]] = { norecurse } +;. diff --git a/llvm/test/Transforms/FunctionAttrs/norecurse_multi_scc_indirect_recursion1.ll b/llvm/test/Transforms/FunctionAttrs/norecurse_multi_scc_indirect_recursion1.ll new file mode 100644 index 0000000..cd94037 --- /dev/null +++ b/llvm/test/Transforms/FunctionAttrs/norecurse_multi_scc_indirect_recursion1.ll @@ -0,0 +1,98 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 5 +; RUN: opt < %s -passes=norecurse-lto-inference -S | FileCheck %s + +; This test includes a call graph with multiple SCCs. The purpose of this is +; to check that norecurse is added to a function which calls a function which +; is indirectly recursive but is not part of the recursive chain. +; There are two SCCs in this test: +; SCC#1: bar2, bar3, bar4 +; SCC#2: baz, fun +; f1() calls bar2 and baz, both of which are part of some indirect recursive +; chain. but does not call back f1() and hence f1() can be marked as +; norecurse. + +define dso_local i32 @main() norecurse { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define dso_local i32 @main( +; CHECK-SAME: ) #[[ATTR0:[0-9]+]] { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @f1() +; CHECK-NEXT: ret i32 0 +; +entry: + tail call void @f1() + ret i32 0 +} + +define internal void @f1() { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define internal void @f1( +; CHECK-SAME: ) #[[ATTR0]] { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar2() +; CHECK-NEXT: tail call void @baz() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar2() + tail call void @baz() + ret void +} + +define internal void @bar4() { +; CHECK-LABEL: define internal void @bar4() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar2() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar2() + ret void +} + +define internal void @bar2() { +; CHECK-LABEL: define internal void @bar2() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar3() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar3() + ret void +} + +define internal void @bar3() { +; CHECK-LABEL: define internal void @bar3() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bar4() +; CHECK-NEXT: ret void +; +entry: + tail call void @bar4() + ret void +} + +define internal void @fun() { +; CHECK-LABEL: define internal void @fun() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @baz() +; CHECK-NEXT: ret void +; +entry: + tail call void @baz() + ret void +} + +define internal void @baz() { +; CHECK-LABEL: define internal void @baz() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @fun() +; CHECK-NEXT: ret void +; +entry: + tail call void @fun() + ret void +} +;. +; CHECK: attributes #[[ATTR0]] = { norecurse } +;. diff --git a/llvm/test/Transforms/FunctionAttrs/norecurse_multinode_refscc.ll b/llvm/test/Transforms/FunctionAttrs/norecurse_multinode_refscc.ll new file mode 100644 index 0000000..8b81a90 --- /dev/null +++ b/llvm/test/Transforms/FunctionAttrs/norecurse_multinode_refscc.ll @@ -0,0 +1,41 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 5 +; RUN: opt -passes=norecurse-lto-inference -S %s | FileCheck %s + +; This is a negative test which results in RefSCC with size > 1. +; RefSCC : [(f2), (f1)] +; --- SCC A (f1) --- size() = 1 +define internal void @f1() { +; CHECK-LABEL: define internal void @f1() { +; CHECK-NEXT: call void @f2() +; CHECK-NEXT: ret void +; + call void @f2() + ret void +} + +; --- SCC B (f2) --- size() = 1 +; f2 indirectly calls f1 using locally allocated function pointer +define internal void @f2() { +; CHECK-LABEL: define internal void @f2() { +; CHECK-NEXT: [[FP:%.*]] = alloca ptr, align 8 +; CHECK-NEXT: store ptr @f1, ptr [[FP]], align 8 +; CHECK-NEXT: [[TMP:%.*]] = load ptr, ptr [[FP]], align 8 +; CHECK-NEXT: call void [[TMP]]() +; CHECK-NEXT: ret void +; + %fp = alloca void ()* + store void ()* @f1, void ()** %fp + %tmp = load void ()*, void ()** %fp + call void %tmp() + ret void +} + +define i32 @main() { +; CHECK-LABEL: define i32 @main() { +; CHECK-NEXT: call void @f1() +; CHECK-NEXT: ret i32 0 +; + call void @f1() + ret i32 0 +} + diff --git a/llvm/test/Transforms/FunctionAttrs/norecurse_self_recursive_callee.ll b/llvm/test/Transforms/FunctionAttrs/norecurse_self_recursive_callee.ll new file mode 100644 index 0000000..461e5df --- /dev/null +++ b/llvm/test/Transforms/FunctionAttrs/norecurse_self_recursive_callee.ll @@ -0,0 +1,88 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals all --version 5 +; RUN: opt < %s -passes=norecurse-lto-inference -S | FileCheck %s + +; This test includes a call graph with a self recursive function. +; The purpose of this is to check that norecurse is added to functions +; which have a self-recursive function in the call-chain. +; The call-chain in this test is as follows +; main -> bob -> callee1 -> callee2 +; where callee2 is self recursive. + +@x = dso_local global i32 4, align 4 +@y = dso_local global i32 2, align 4 + +;. +; CHECK: @x = dso_local global i32 4, align 4 +; CHECK: @y = dso_local global i32 2, align 4 +;. +define internal void @callee2() { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define internal void @callee2( +; CHECK-SAME: ) #[[ATTR0:[0-9]+]] { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, ptr @y, align 4 +; CHECK-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 +; CHECK-NEXT: store volatile i32 [[INC]], ptr @y, align 4 +; CHECK-NEXT: ret void +; +entry: + %0 = load volatile i32, ptr @y, align 4 + %inc = add nsw i32 %0, 1 + store volatile i32 %inc, ptr @y, align 4 + ret void +} + +define internal void @callee1(i32 %x) { +; CHECK-LABEL: define internal void @callee1( +; CHECK-SAME: i32 [[X:%.*]]) { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 [[X]], 0 +; CHECK-NEXT: br i1 [[CMP]], label %[[IF_THEN:.*]], label %[[IF_END:.*]] +; CHECK: [[IF_THEN]]: +; CHECK-NEXT: tail call void @callee1(i32 [[X]]) +; CHECK-NEXT: br label %[[IF_END]] +; CHECK: [[IF_END]]: +; CHECK-NEXT: tail call void @callee2() +; CHECK-NEXT: ret void +; +entry: + %cmp = icmp sgt i32 %x, 0 + br i1 %cmp, label %if.then, label %if.end + +if.then: ; preds = %entry + tail call void @callee1(i32 %x) + br label %if.end + +if.end: ; preds = %if.then, %entry + tail call void @callee2() + ret void +} + +define internal void @bob() { +; CHECK-LABEL: define internal void @bob() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, ptr @x, align 4 +; CHECK-NEXT: tail call void @callee2(i32 [[TMP0]]) +; CHECK-NEXT: ret void +; +entry: + %0 = load volatile i32, ptr @x, align 4 + tail call void @callee2(i32 %0) + ret void +} + +define dso_local i32 @main() norecurse { +; CHECK: Function Attrs: norecurse +; CHECK-LABEL: define dso_local i32 @main( +; CHECK-SAME: ) #[[ATTR0]] { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: tail call void @bob() +; CHECK-NEXT: ret i32 0 +; +entry: + tail call void @bob() + ret i32 0 +} +;. +; CHECK: attributes #[[ATTR0]] = { norecurse } +;. diff --git a/llvm/test/Transforms/InstCombine/select-safe-bool-transforms.ll b/llvm/test/Transforms/InstCombine/select-safe-bool-transforms.ll index 9de9150..8b0a5ca 100644 --- a/llvm/test/Transforms/InstCombine/select-safe-bool-transforms.ll +++ b/llvm/test/Transforms/InstCombine/select-safe-bool-transforms.ll @@ -1,4 +1,4 @@ -; NOTE: Assertions have been autogenerated by utils/update_test_checks.py +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals ; RUN: opt < %s -passes=instcombine -S | FileCheck %s ; TODO: All of these should be optimized to less than or equal to a single @@ -7,13 +7,13 @@ ; --- (A op B) op' A / (B op A) op' A --- ; (A land B) land A -define i1 @land_land_left1(i1 %A, i1 %B) { +define i1 @land_land_left1(i1 %A, i1 %B) !prof !0 { ; CHECK-LABEL: @land_land_left1( -; CHECK-NEXT: [[C:%.*]] = select i1 [[A:%.*]], i1 [[B:%.*]], i1 false +; CHECK-NEXT: [[C:%.*]] = select i1 [[A:%.*]], i1 [[B:%.*]], i1 false, !prof [[PROF1:![0-9]+]] ; CHECK-NEXT: ret i1 [[C]] ; - %c = select i1 %A, i1 %B, i1 false - %res = select i1 %c, i1 %A, i1 false + %c = select i1 %A, i1 %B, i1 false, !prof !1 + %res = select i1 %c, i1 %A, i1 false, !prof !2 ret i1 %res } define i1 @land_land_left2(i1 %A, i1 %B) { @@ -157,13 +157,13 @@ define i1 @lor_band_left2(i1 %A, i1 %B) { } ; (A lor B) lor A -define i1 @lor_lor_left1(i1 %A, i1 %B) { +define i1 @lor_lor_left1(i1 %A, i1 %B) !prof !0 { ; CHECK-LABEL: @lor_lor_left1( -; CHECK-NEXT: [[C:%.*]] = select i1 [[A:%.*]], i1 true, i1 [[B:%.*]] +; CHECK-NEXT: [[C:%.*]] = select i1 [[A:%.*]], i1 true, i1 [[B:%.*]], !prof [[PROF1]] ; CHECK-NEXT: ret i1 [[C]] ; - %c = select i1 %A, i1 true, i1 %B - %res = select i1 %c, i1 true, i1 %A + %c = select i1 %A, i1 true, i1 %B, !prof !1 + %res = select i1 %c, i1 true, i1 %A, !prof !2 ret i1 %res } define i1 @lor_lor_left2(i1 %A, i1 %B) { @@ -506,3 +506,12 @@ define <2 x i1> @PR50500_falseval(<2 x i1> %a, <2 x i1> %b) { %r = select <2 x i1> %a, <2 x i1> %b, <2 x i1> %s ret <2 x i1> %r } + +!0 = !{!"function_entry_count", i64 1000} +!1 = !{!"branch_weights", i32 2, i32 3} +!2 = !{!"branch_weights", i32 5, i32 7} + +;. +; CHECK: [[META0:![0-9]+]] = !{!"function_entry_count", i64 1000} +; CHECK: [[PROF1]] = !{!"branch_weights", i32 2, i32 3} +;. diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/neon-inloop-reductions.ll b/llvm/test/Transforms/LoopVectorize/AArch64/neon-inloop-reductions.ll new file mode 100644 index 0000000..22696d0 --- /dev/null +++ b/llvm/test/Transforms/LoopVectorize/AArch64/neon-inloop-reductions.ll @@ -0,0 +1,121 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6 +; RUN: opt -p loop-vectorize -prefer-inloop-reductions -mcpu=apple-m1 -force-vector-interleave=1 -S %s | FileCheck %s + +target triple = "arm64-apple-macosx" + +define i32 @mul_used_outside_vpexpression(ptr %src.0, ptr %src.1) { +; CHECK-LABEL: define i32 @mul_used_outside_vpexpression( +; CHECK-SAME: ptr [[SRC_0:%.*]], ptr [[SRC_1:%.*]]) #[[ATTR0:[0-9]+]] { +; CHECK-NEXT: [[ITER_CHECK:.*]]: +; CHECK-NEXT: br i1 false, label %[[VEC_EPILOG_SCALAR_PH:.*]], label %[[VECTOR_MAIN_LOOP_ITER_CHECK:.*]] +; CHECK: [[VECTOR_MAIN_LOOP_ITER_CHECK]]: +; CHECK-NEXT: br i1 false, label %[[VEC_EPILOG_PH:.*]], label %[[VECTOR_PH:.*]] +; CHECK: [[VECTOR_PH]]: +; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i8, ptr [[SRC_1]], i64 1 +; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] +; CHECK: [[VECTOR_BODY]]: +; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; CHECK-NEXT: [[VEC_PHI:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ] +; CHECK-NEXT: [[VEC_PHI1:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[TMP8:%.*]], %[[VECTOR_BODY]] ] +; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[SRC_0]], i64 [[INDEX]] +; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[NEXT_GEP]], align 1 +; CHECK-NEXT: [[TMP1:%.*]] = load i8, ptr [[TMP0]], align 1 +; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <16 x i8> poison, i8 [[TMP1]], i64 0 +; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <16 x i8> [[BROADCAST_SPLATINSERT]], <16 x i8> poison, <16 x i32> zeroinitializer +; CHECK-NEXT: [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32> +; CHECK-NEXT: [[TMP3:%.*]] = zext <16 x i8> [[BROADCAST_SPLAT]] to <16 x i32> +; CHECK-NEXT: [[TMP4:%.*]] = mul <16 x i32> [[TMP2]], [[TMP3]] +; CHECK-NEXT: [[TMP5:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP4]]) +; CHECK-NEXT: [[TMP6]] = add i32 [[VEC_PHI]], [[TMP5]] +; CHECK-NEXT: [[TMP7:%.*]] = call i32 @llvm.vector.reduce.or.v16i32(<16 x i32> [[TMP4]]) +; CHECK-NEXT: [[TMP8]] = or i32 [[VEC_PHI1]], [[TMP7]] +; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16 +; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96 +; CHECK-NEXT: br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; CHECK: [[MIDDLE_BLOCK]]: +; CHECK-NEXT: br i1 false, label %[[EXIT:.*]], label %[[VEC_EPILOG_ITER_CHECK:.*]] +; CHECK: [[VEC_EPILOG_ITER_CHECK]]: +; CHECK-NEXT: [[IND_END:%.*]] = getelementptr i8, ptr [[SRC_0]], i64 96 +; CHECK-NEXT: br i1 false, label %[[VEC_EPILOG_SCALAR_PH]], label %[[VEC_EPILOG_PH]], !prof [[PROF3:![0-9]+]] +; CHECK: [[VEC_EPILOG_PH]]: +; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ 96, %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[VECTOR_MAIN_LOOP_ITER_CHECK]] ] +; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[VECTOR_MAIN_LOOP_ITER_CHECK]] ] +; CHECK-NEXT: [[BC_MERGE_RDX2:%.*]] = phi i32 [ [[TMP8]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[VECTOR_MAIN_LOOP_ITER_CHECK]] ] +; CHECK-NEXT: [[TMP10:%.*]] = getelementptr i8, ptr [[SRC_0]], i64 100 +; CHECK-NEXT: [[TMP11:%.*]] = getelementptr i8, ptr [[SRC_1]], i64 1 +; CHECK-NEXT: br label %[[VEC_EPILOG_VECTOR_BODY:.*]] +; CHECK: [[VEC_EPILOG_VECTOR_BODY]]: +; CHECK-NEXT: [[INDEX3:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT10:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ] +; CHECK-NEXT: [[VEC_PHI4:%.*]] = phi i32 [ [[BC_MERGE_RDX]], %[[VEC_EPILOG_PH]] ], [ [[TMP17:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ] +; CHECK-NEXT: [[VEC_PHI5:%.*]] = phi i32 [ [[BC_MERGE_RDX2]], %[[VEC_EPILOG_PH]] ], [ [[TMP19:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ] +; CHECK-NEXT: [[NEXT_GEP6:%.*]] = getelementptr i8, ptr [[SRC_0]], i64 [[INDEX3]] +; CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[NEXT_GEP6]], align 1 +; CHECK-NEXT: [[TMP12:%.*]] = load i8, ptr [[TMP11]], align 1 +; CHECK-NEXT: [[BROADCAST_SPLATINSERT8:%.*]] = insertelement <4 x i8> poison, i8 [[TMP12]], i64 0 +; CHECK-NEXT: [[BROADCAST_SPLAT9:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT8]], <4 x i8> poison, <4 x i32> zeroinitializer +; CHECK-NEXT: [[TMP13:%.*]] = zext <4 x i8> [[WIDE_LOAD7]] to <4 x i32> +; CHECK-NEXT: [[TMP14:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT9]] to <4 x i32> +; CHECK-NEXT: [[TMP15:%.*]] = mul <4 x i32> [[TMP13]], [[TMP14]] +; CHECK-NEXT: [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP15]]) +; CHECK-NEXT: [[TMP17]] = add i32 [[VEC_PHI4]], [[TMP16]] +; CHECK-NEXT: [[TMP18:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP15]]) +; CHECK-NEXT: [[TMP19]] = or i32 [[VEC_PHI5]], [[TMP18]] +; CHECK-NEXT: [[INDEX_NEXT10]] = add nuw i64 [[INDEX3]], 4 +; CHECK-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT10]], 100 +; CHECK-NEXT: br i1 [[TMP20]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] +; CHECK: [[VEC_EPILOG_MIDDLE_BLOCK]]: +; CHECK-NEXT: br i1 false, label %[[EXIT]], label %[[VEC_EPILOG_SCALAR_PH]] +; CHECK: [[VEC_EPILOG_SCALAR_PH]]: +; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 100, %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ 96, %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ] +; CHECK-NEXT: [[BC_RESUME_VAL11:%.*]] = phi ptr [ [[TMP10]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END]], %[[VEC_EPILOG_ITER_CHECK]] ], [ [[SRC_0]], %[[ITER_CHECK]] ] +; CHECK-NEXT: [[BC_MERGE_RDX12:%.*]] = phi i32 [ [[TMP17]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[TMP6]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ] +; CHECK-NEXT: [[BC_MERGE_RDX13:%.*]] = phi i32 [ [[TMP19]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[TMP8]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ] +; CHECK-NEXT: br label %[[LOOP:.*]] +; CHECK: [[LOOP]]: +; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; CHECK-NEXT: [[PTR_IV:%.*]] = phi ptr [ [[BC_RESUME_VAL11]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[GEP_0:%.*]], %[[LOOP]] ] +; CHECK-NEXT: [[RED_0:%.*]] = phi i32 [ [[BC_MERGE_RDX12]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[RED_0_NEXT:%.*]], %[[LOOP]] ] +; CHECK-NEXT: [[RED_1:%.*]] = phi i32 [ [[BC_MERGE_RDX13]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[RED_1_NEXT:%.*]], %[[LOOP]] ] +; CHECK-NEXT: [[GEP_0]] = getelementptr i8, ptr [[PTR_IV]], i64 1 +; CHECK-NEXT: [[L_0:%.*]] = load i8, ptr [[PTR_IV]], align 1 +; CHECK-NEXT: [[GEP_1:%.*]] = getelementptr i8, ptr [[SRC_1]], i64 1 +; CHECK-NEXT: [[L_1:%.*]] = load i8, ptr [[GEP_1]], align 1 +; CHECK-NEXT: [[L_0_EXT:%.*]] = zext i8 [[L_0]] to i32 +; CHECK-NEXT: [[L_1_EXT:%.*]] = zext i8 [[L_1]] to i32 +; CHECK-NEXT: [[MUL_EXT_LL:%.*]] = mul i32 [[L_0_EXT]], [[L_1_EXT]] +; CHECK-NEXT: [[RED_1_NEXT]] = or i32 [[MUL_EXT_LL]], [[RED_1]] +; CHECK-NEXT: [[RED_0_NEXT]] = add i32 [[MUL_EXT_LL]], [[RED_0]] +; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1 +; CHECK-NEXT: [[EC:%.*]] = icmp eq i32 [[IV]], 101 +; CHECK-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]] +; CHECK: [[EXIT]]: +; CHECK-NEXT: [[RED_1_NEXT_LCSSA:%.*]] = phi i32 [ [[RED_1_NEXT]], %[[LOOP]] ], [ [[TMP8]], %[[MIDDLE_BLOCK]] ], [ [[TMP19]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ] +; CHECK-NEXT: [[RED_0_NEXT_LCSSA:%.*]] = phi i32 [ [[RED_0_NEXT]], %[[LOOP]] ], [ [[TMP6]], %[[MIDDLE_BLOCK]] ], [ [[TMP17]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ] +; CHECK-NEXT: [[RES:%.*]] = add i32 [[RED_1_NEXT_LCSSA]], [[RED_0_NEXT_LCSSA]] +; CHECK-NEXT: ret i32 [[RES]] +; +entry: + br label %loop + +loop: + %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ] + %ptr.iv = phi ptr [ %src.0, %entry ], [ %gep.0, %loop ] + %red.0 = phi i32 [ 0, %entry ], [ %red.0.next, %loop ] + %red.1 = phi i32 [ 0, %entry ], [ %red.1.next, %loop ] + %gep.0 = getelementptr i8, ptr %ptr.iv, i64 1 + %l.0 = load i8, ptr %ptr.iv, align 1 + %gep.1 = getelementptr i8, ptr %src.1, i64 1 + %l.1 = load i8, ptr %gep.1, align 1 + %l.0.ext = zext i8 %l.0 to i32 + %l.1.ext = zext i8 %l.1 to i32 + %mul.ext.ll = mul i32 %l.0.ext, %l.1.ext + %red.1.next = or i32 %mul.ext.ll, %red.1 + %red.0.next = add i32 %mul.ext.ll, %red.0 + %iv.next = add i32 %iv, 1 + %ec = icmp eq i32 %iv, 101 + br i1 %ec, label %exit, label %loop + +exit: + %res = add i32 %red.1.next, %red.0.next + ret i32 %res +} diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/pr162009.ll b/llvm/test/Transforms/LoopVectorize/AArch64/pr162009.ll new file mode 100644 index 0000000..6095b24 --- /dev/null +++ b/llvm/test/Transforms/LoopVectorize/AArch64/pr162009.ll @@ -0,0 +1,79 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 +; RUN: opt -passes=loop-vectorize -force-vector-interleave=1 -enable-epilogue-vectorization=false -S < %s | FileCheck %s --check-prefixes=CHECK-NO-PARTIAL-REDUCTION + +target triple = "aarch64" + +define i128 @add_reduc_i32_i128_unsupported(ptr %a, ptr %b) "target-features"="+dotprod" { +; CHECK-NO-PARTIAL-REDUCTION-LABEL: define i128 @add_reduc_i32_i128_unsupported( +; CHECK-NO-PARTIAL-REDUCTION-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) #[[ATTR0:[0-9]+]] { +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[ENTRY:.*:]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: br label %[[VECTOR_PH:.*]] +; CHECK-NO-PARTIAL-REDUCTION: [[VECTOR_PH]]: +; CHECK-NO-PARTIAL-REDUCTION-NEXT: br label %[[VECTOR_BODY:.*]] +; CHECK-NO-PARTIAL-REDUCTION: [[VECTOR_BODY]]: +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[VEC_PHI:%.*]] = phi <4 x i128> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP0:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 1 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP1:%.*]] = zext <4 x i32> [[WIDE_LOAD]] to <4 x i64> +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP2:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 1 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP3:%.*]] = zext <4 x i32> [[WIDE_LOAD1]] to <4 x i64> +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP4:%.*]] = mul nuw <4 x i64> [[TMP1]], [[TMP3]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP5:%.*]] = zext <4 x i64> [[TMP4]] to <4 x i128> +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP7]] = add <4 x i128> [[VEC_PHI]], [[TMP5]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4024 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; CHECK-NO-PARTIAL-REDUCTION: [[MIDDLE_BLOCK]]: +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[TMP8:%.*]] = call i128 @llvm.vector.reduce.add.v4i128(<4 x i128> [[TMP7]]) +; CHECK-NO-PARTIAL-REDUCTION-NEXT: br label %[[SCALAR_PH:.*]] +; CHECK-NO-PARTIAL-REDUCTION: [[SCALAR_PH]]: +; CHECK-NO-PARTIAL-REDUCTION-NEXT: br label %[[FOR_BODY:.*]] +; CHECK-NO-PARTIAL-REDUCTION: [[FOR_BODY]]: +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[IV:%.*]] = phi i64 [ 4024, %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[FOR_BODY]] ] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[ACCUM:%.*]] = phi i128 [ [[TMP8]], %[[SCALAR_PH]] ], [ [[ADD:%.*]], %[[FOR_BODY]] ] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[GEP_A:%.*]] = getelementptr i32, ptr [[A]], i64 [[IV]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[LOAD_A:%.*]] = load i32, ptr [[GEP_A]], align 1 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[EXT_A:%.*]] = zext i32 [[LOAD_A]] to i64 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[GEP_B:%.*]] = getelementptr i32, ptr [[B]], i64 [[IV]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[LOAD_B:%.*]] = load i32, ptr [[GEP_B]], align 1 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[EXT_B:%.*]] = zext i32 [[LOAD_B]] to i64 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[MUL:%.*]] = mul nuw i64 [[EXT_A]], [[EXT_B]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[MUL_ZEXT:%.*]] = zext i64 [[MUL]] to i128 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[ADD]] = add i128 [[ACCUM]], [[MUL_ZEXT]] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 4025 +; CHECK-NO-PARTIAL-REDUCTION-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_EXIT:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]] +; CHECK-NO-PARTIAL-REDUCTION: [[FOR_EXIT]]: +; CHECK-NO-PARTIAL-REDUCTION-NEXT: [[ADD_LCSSA:%.*]] = phi i128 [ [[ADD]], %[[FOR_BODY]] ] +; CHECK-NO-PARTIAL-REDUCTION-NEXT: ret i128 [[ADD_LCSSA]] +; +entry: + br label %for.body + +for.body: + %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ] + %accum = phi i128 [ 0, %entry ], [ %add, %for.body ] + %gep.a = getelementptr i32, ptr %a, i64 %iv + %load.a = load i32, ptr %gep.a, align 1 + %ext.a = zext i32 %load.a to i64 + %gep.b = getelementptr i32, ptr %b, i64 %iv + %load.b = load i32, ptr %gep.b, align 1 + %ext.b = zext i32 %load.b to i64 + %mul = mul nuw i64 %ext.a, %ext.b + %mul.zext = zext i64 %mul to i128 + %add = add i128 %accum, %mul.zext + %iv.next = add i64 %iv, 1 + %exitcond.not = icmp eq i64 %iv.next, 4025 + br i1 %exitcond.not, label %for.exit, label %for.body + +for.exit: + ret i128 %add +} +;. +; CHECK-NO-PARTIAL-REDUCTION: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} +; CHECK-NO-PARTIAL-REDUCTION: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} +; CHECK-NO-PARTIAL-REDUCTION: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} +; CHECK-NO-PARTIAL-REDUCTION: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]} +;. diff --git a/llvm/test/Transforms/LoopVectorize/ARM/replicating-load-store-costs.ll b/llvm/test/Transforms/LoopVectorize/ARM/replicating-load-store-costs.ll new file mode 100644 index 0000000..fd83a01 --- /dev/null +++ b/llvm/test/Transforms/LoopVectorize/ARM/replicating-load-store-costs.ll @@ -0,0 +1,84 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 +; RUN: opt -p loop-vectorize -S %s | FileCheck %s + +target triple = "armv7-unknown-linux-gnueabihf" + +define void @replicating_load_used_by_other_load(i32 %arg, ptr %a, i32 %b) { +; CHECK-LABEL: define void @replicating_load_used_by_other_load( +; CHECK-SAME: i32 [[ARG:%.*]], ptr [[A:%.*]], i32 [[B:%.*]]) { +; CHECK-NEXT: [[ENTRY:.*]]: +; CHECK-NEXT: br label %[[LOOP:.*]] +; CHECK: [[LOOP]]: +; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[ARG]], %[[ENTRY]] ] +; CHECK-NEXT: [[SHR:%.*]] = lshr i32 [[IV]], 1 +; CHECK-NEXT: [[AND_1:%.*]] = and i32 [[IV]], 1 +; CHECK-NEXT: [[SHL_1:%.*]] = shl i32 [[IV]], 2 +; CHECK-NEXT: [[SHL_2:%.*]] = shl i32 [[IV]], 1 +; CHECK-NEXT: [[AND_2:%.*]] = and i32 [[SHL_2]], 2 +; CHECK-NEXT: [[OR_1:%.*]] = or i32 [[AND_2]], [[AND_1]] +; CHECK-NEXT: [[OR_2:%.*]] = or i32 [[OR_1]], [[SHL_1]] +; CHECK-NEXT: [[XOR_1:%.*]] = xor i32 [[B]], [[OR_2]] +; CHECK-NEXT: [[XOR_2:%.*]] = xor i32 [[XOR_1]], [[ARG]] +; CHECK-NEXT: [[SHR_2:%.*]] = lshr i32 [[SHL_1]], 1 +; CHECK-NEXT: [[XOR_3:%.*]] = xor i32 [[SHR]], [[ARG]] +; CHECK-NEXT: [[AND_3:%.*]] = and i32 [[XOR_3]], 1 +; CHECK-NEXT: [[AND_4:%.*]] = and i32 [[IV]], 2147483646 +; CHECK-NEXT: [[OR_3:%.*]] = or i32 [[AND_3]], [[AND_4]] +; CHECK-NEXT: [[AND_5:%.*]] = and i32 [[IV]], 254 +; CHECK-NEXT: [[SHL_3:%.*]] = shl i32 [[OR_3]], 1 +; CHECK-NEXT: [[XOR_4:%.*]] = xor i32 [[SHL_3]], 2 +; CHECK-NEXT: [[OR_4:%.*]] = or i32 [[AND_5]], [[XOR_4]] +; CHECK-NEXT: [[XOR_5:%.*]] = xor i32 [[SHR_2]], [[OR_4]] +; CHECK-NEXT: [[XOR_6:%.*]] = xor i32 [[XOR_5]], [[XOR_2]] +; CHECK-NEXT: [[AND_6:%.*]] = and i32 [[XOR_6]], 255 +; CHECK-NEXT: [[XOR_7:%.*]] = xor i32 [[AND_6]], 1 +; CHECK-NEXT: [[GEP:%.*]] = getelementptr i8, ptr [[A]], i32 [[XOR_7]] +; CHECK-NEXT: [[LD:%.*]] = load i8, ptr [[GEP]], align 1 +; CHECK-NEXT: [[ZEXT:%.*]] = zext i8 [[LD]] to i32 +; CHECK-NEXT: [[GEP_2:%.*]] = getelementptr i32, ptr null, i32 [[ZEXT]] +; CHECK-NEXT: store i32 0, ptr [[GEP_2]], align 4 +; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1 +; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[IV_NEXT]], 100 +; CHECK-NEXT: br i1 [[CMP]], label %[[EXIT:.*]], label %[[LOOP]] +; CHECK: [[EXIT]]: +; CHECK-NEXT: ret void +; +entry: + br label %loop + +loop: + %iv = phi i32 [ %iv.next, %loop ], [ %arg, %entry ] + %shr = lshr i32 %iv, 1 + %and.1 = and i32 %iv, 1 + %shl.1 = shl i32 %iv, 2 + %shl.2 = shl i32 %iv, 1 + %and.2 = and i32 %shl.2, 2 + %or.1 = or i32 %and.2, %and.1 + %or.2 = or i32 %or.1, %shl.1 + %xor.1 = xor i32 %b, %or.2 + %xor.2 = xor i32 %xor.1, %arg + %shr.2 = lshr i32 %shl.1, 1 + %xor.3 = xor i32 %shr, %arg + %and.3 = and i32 %xor.3, 1 + %and.4 = and i32 %iv, 2147483646 + %or.3 = or i32 %and.3, %and.4 + %and.5 = and i32 %iv, 254 + %shl.3 = shl i32 %or.3, 1 + %xor.4 = xor i32 %shl.3, 2 + %or.4 = or i32 %and.5, %xor.4 + %xor.5 = xor i32 %shr.2, %or.4 + %xor.6 = xor i32 %xor.5, %xor.2 + %and.6 = and i32 %xor.6, 255 + %xor.7 = xor i32 %and.6, 1 + %gep = getelementptr i8, ptr %a, i32 %xor.7 + %ld = load i8, ptr %gep, align 1 + %zext = zext i8 %ld to i32 + %gep.2 = getelementptr i32, ptr null, i32 %zext + store i32 0, ptr %gep.2, align 4 + %iv.next = add i32 %iv, 1 + %cmp = icmp eq i32 %iv.next, 100 + br i1 %cmp, label %exit, label %loop + +exit: + ret void +} diff --git a/llvm/test/Transforms/LoopVectorize/X86/replicating-load-store-costs.ll b/llvm/test/Transforms/LoopVectorize/X86/replicating-load-store-costs.ll index 8784873..f5329cf 100644 --- a/llvm/test/Transforms/LoopVectorize/X86/replicating-load-store-costs.ll +++ b/llvm/test/Transforms/LoopVectorize/X86/replicating-load-store-costs.ll @@ -454,6 +454,132 @@ exit: ret void } +declare i1 @cond() + +define double @test_load_used_by_other_load_scev(ptr %ptr.a, ptr %ptr.b, ptr %ptr.c) { +; I64-LABEL: define double @test_load_used_by_other_load_scev( +; I64-SAME: ptr [[PTR_A:%.*]], ptr [[PTR_B:%.*]], ptr [[PTR_C:%.*]]) { +; I64-NEXT: [[ENTRY:.*]]: +; I64-NEXT: br label %[[OUTER_LOOP:.*]] +; I64: [[OUTER_LOOP_LOOPEXIT:.*]]: +; I64-NEXT: br label %[[OUTER_LOOP]] +; I64: [[OUTER_LOOP]]: +; I64-NEXT: [[ACCUM:%.*]] = phi double [ 0.000000e+00, %[[ENTRY]] ], [ [[TMP29:%.*]], %[[OUTER_LOOP_LOOPEXIT]] ] +; I64-NEXT: [[COND:%.*]] = call i1 @cond() +; I64-NEXT: br i1 [[COND]], label %[[INNER_LOOP_PREHEADER:.*]], label %[[EXIT:.*]] +; I64: [[INNER_LOOP_PREHEADER]]: +; I64-NEXT: br label %[[VECTOR_PH:.*]] +; I64: [[VECTOR_PH]]: +; I64-NEXT: br label %[[VECTOR_BODY:.*]] +; I64: [[VECTOR_BODY]]: +; I64-NEXT: [[TMP0:%.*]] = add i64 0, 1 +; I64-NEXT: [[TMP1:%.*]] = add i64 1, 1 +; I64-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[PTR_C]], i64 [[TMP0]] +; I64-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[PTR_C]], i64 [[TMP1]] +; I64-NEXT: [[TMP4:%.*]] = getelementptr i64, ptr [[PTR_A]], i64 [[TMP0]] +; I64-NEXT: [[TMP5:%.*]] = getelementptr i64, ptr [[PTR_A]], i64 [[TMP1]] +; I64-NEXT: [[TMP6:%.*]] = load i64, ptr [[TMP4]], align 8 +; I64-NEXT: [[TMP7:%.*]] = load i64, ptr [[TMP5]], align 8 +; I64-NEXT: [[TMP8:%.*]] = getelementptr double, ptr [[PTR_B]], i64 [[TMP6]] +; I64-NEXT: [[TMP9:%.*]] = getelementptr double, ptr [[PTR_B]], i64 [[TMP7]] +; I64-NEXT: [[TMP10:%.*]] = load double, ptr [[PTR_A]], align 8 +; I64-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x double> poison, double [[TMP10]], i64 0 +; I64-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x double> [[BROADCAST_SPLATINSERT]], <2 x double> poison, <2 x i32> zeroinitializer +; I64-NEXT: [[TMP11:%.*]] = fadd <2 x double> [[BROADCAST_SPLAT]], zeroinitializer +; I64-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[TMP2]], i64 8 +; I64-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[TMP3]], i64 8 +; I64-NEXT: [[TMP14:%.*]] = load double, ptr [[TMP12]], align 8 +; I64-NEXT: [[TMP15:%.*]] = load double, ptr [[TMP13]], align 8 +; I64-NEXT: [[TMP16:%.*]] = insertelement <2 x double> poison, double [[TMP14]], i32 0 +; I64-NEXT: [[TMP17:%.*]] = insertelement <2 x double> [[TMP16]], double [[TMP15]], i32 1 +; I64-NEXT: [[TMP18:%.*]] = fmul <2 x double> [[TMP11]], zeroinitializer +; I64-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <2 x double> poison, double [[ACCUM]], i64 0 +; I64-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <2 x double> [[BROADCAST_SPLATINSERT1]], <2 x double> poison, <2 x i32> zeroinitializer +; I64-NEXT: [[TMP19:%.*]] = shufflevector <2 x double> [[BROADCAST_SPLAT2]], <2 x double> [[TMP18]], <2 x i32> <i32 1, i32 2> +; I64-NEXT: [[TMP20:%.*]] = fmul <2 x double> [[TMP17]], zeroinitializer +; I64-NEXT: [[TMP21:%.*]] = fadd <2 x double> [[TMP20]], zeroinitializer +; I64-NEXT: [[TMP22:%.*]] = fadd <2 x double> [[TMP21]], splat (double 1.000000e+00) +; I64-NEXT: [[TMP23:%.*]] = load double, ptr [[TMP8]], align 8 +; I64-NEXT: [[TMP24:%.*]] = load double, ptr [[TMP9]], align 8 +; I64-NEXT: [[TMP25:%.*]] = insertelement <2 x double> poison, double [[TMP23]], i32 0 +; I64-NEXT: [[TMP26:%.*]] = insertelement <2 x double> [[TMP25]], double [[TMP24]], i32 1 +; I64-NEXT: [[TMP27:%.*]] = fdiv <2 x double> [[TMP26]], [[TMP22]] +; I64-NEXT: [[TMP28:%.*]] = fsub <2 x double> [[TMP19]], [[TMP27]] +; I64-NEXT: br label %[[MIDDLE_BLOCK:.*]] +; I64: [[MIDDLE_BLOCK]]: +; I64-NEXT: [[TMP29]] = extractelement <2 x double> [[TMP28]], i32 1 +; I64-NEXT: br label %[[OUTER_LOOP_LOOPEXIT]] +; I64: [[EXIT]]: +; I64-NEXT: ret double [[ACCUM]] +; +; I32-LABEL: define double @test_load_used_by_other_load_scev( +; I32-SAME: ptr [[PTR_A:%.*]], ptr [[PTR_B:%.*]], ptr [[PTR_C:%.*]]) { +; I32-NEXT: [[ENTRY:.*]]: +; I32-NEXT: br label %[[OUTER_LOOP:.*]] +; I32: [[OUTER_LOOP]]: +; I32-NEXT: [[ACCUM:%.*]] = phi double [ 0.000000e+00, %[[ENTRY]] ], [ [[RESULT:%.*]], %[[INNER_LOOP:.*]] ] +; I32-NEXT: [[COND:%.*]] = call i1 @cond() +; I32-NEXT: br i1 [[COND]], label %[[INNER_LOOP]], label %[[EXIT:.*]] +; I32: [[INNER_LOOP]]: +; I32-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[OUTER_LOOP]] ], [ [[IV_NEXT:%.*]], %[[INNER_LOOP]] ] +; I32-NEXT: [[ACCUM_INNER:%.*]] = phi double [ [[ACCUM]], %[[OUTER_LOOP]] ], [ [[MUL1:%.*]], %[[INNER_LOOP]] ] +; I32-NEXT: [[IDX_PLUS1:%.*]] = add i64 [[IV]], 1 +; I32-NEXT: [[GEP_C:%.*]] = getelementptr i8, ptr [[PTR_C]], i64 [[IDX_PLUS1]] +; I32-NEXT: [[GEP_A_I64:%.*]] = getelementptr i64, ptr [[PTR_A]], i64 [[IDX_PLUS1]] +; I32-NEXT: [[LOAD_IDX:%.*]] = load i64, ptr [[GEP_A_I64]], align 8 +; I32-NEXT: [[GEP_B:%.*]] = getelementptr double, ptr [[PTR_B]], i64 [[LOAD_IDX]] +; I32-NEXT: [[LOAD_A:%.*]] = load double, ptr [[PTR_A]], align 8 +; I32-NEXT: [[ADD1:%.*]] = fadd double [[LOAD_A]], 0.000000e+00 +; I32-NEXT: [[GEP_C_OFFSET:%.*]] = getelementptr i8, ptr [[GEP_C]], i64 8 +; I32-NEXT: [[LOAD_C:%.*]] = load double, ptr [[GEP_C_OFFSET]], align 8 +; I32-NEXT: [[MUL1]] = fmul double [[ADD1]], 0.000000e+00 +; I32-NEXT: [[MUL2:%.*]] = fmul double [[LOAD_C]], 0.000000e+00 +; I32-NEXT: [[ADD2:%.*]] = fadd double [[MUL2]], 0.000000e+00 +; I32-NEXT: [[ADD3:%.*]] = fadd double [[ADD2]], 1.000000e+00 +; I32-NEXT: [[LOAD_B:%.*]] = load double, ptr [[GEP_B]], align 8 +; I32-NEXT: [[DIV:%.*]] = fdiv double [[LOAD_B]], [[ADD3]] +; I32-NEXT: [[RESULT]] = fsub double [[ACCUM_INNER]], [[DIV]] +; I32-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 +; I32-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 1 +; I32-NEXT: br i1 [[EXITCOND]], label %[[OUTER_LOOP]], label %[[INNER_LOOP]] +; I32: [[EXIT]]: +; I32-NEXT: ret double [[ACCUM]] +; +entry: + br label %outer.loop + +outer.loop: + %accum = phi double [ 0.0, %entry ], [ %result, %inner.loop ] + %cond = call i1 @cond() + br i1 %cond, label %inner.loop, label %exit + +inner.loop: + %iv = phi i64 [ 0, %outer.loop ], [ %iv.next, %inner.loop ] + %accum.inner = phi double [ %accum, %outer.loop ], [ %mul1, %inner.loop ] + %idx.plus1 = add i64 %iv, 1 + %gep.c = getelementptr i8, ptr %ptr.c, i64 %idx.plus1 + %gep.a.i64 = getelementptr i64, ptr %ptr.a, i64 %idx.plus1 + %load.idx = load i64, ptr %gep.a.i64, align 8 + %gep.b = getelementptr double, ptr %ptr.b, i64 %load.idx + %load.a = load double, ptr %ptr.a, align 8 + %add1 = fadd double %load.a, 0.000000e+00 + %gep.c.offset = getelementptr i8, ptr %gep.c, i64 8 + %load.c = load double, ptr %gep.c.offset, align 8 + %mul1 = fmul double %add1, 0.000000e+00 + %mul2 = fmul double %load.c, 0.000000e+00 + %add2 = fadd double %mul2, 0.000000e+00 + %add3 = fadd double %add2, 1.000000e+00 + %load.b = load double, ptr %gep.b, align 8 + %div = fdiv double %load.b, %add3 + %result = fsub double %accum.inner, %div + %iv.next = add i64 %iv, 1 + %exitcond = icmp eq i64 %iv, 1 + br i1 %exitcond, label %outer.loop, label %inner.loop + +exit: + ret double %accum +} + attributes #0 = { "target-cpu"="znver2" } !0 = distinct !{!0, !1} diff --git a/llvm/test/Transforms/SCCP/relax-range-checks.ll b/llvm/test/Transforms/SCCP/relax-range-checks.ll index 90722f3..34e4813 100644 --- a/llvm/test/Transforms/SCCP/relax-range-checks.ll +++ b/llvm/test/Transforms/SCCP/relax-range-checks.ll @@ -89,4 +89,28 @@ define i1 @relax_range_check_multiuse(i8 range(i8 0, 5) %x) { ret i1 %ret } +define i1 @range_check_to_icmp_eq1(i32 range(i32 0, 4) %x) { +; CHECK-LABEL: define i1 @range_check_to_icmp_eq1( +; CHECK-SAME: i32 range(i32 0, 4) [[X:%.*]]) { +; CHECK-NEXT: [[OFF:%.*]] = add nsw i32 [[X]], -3 +; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[X]], 3 +; CHECK-NEXT: ret i1 [[TMP1]] +; + %off = add nsw i32 %x, -3 + %cmp = icmp ult i32 %off, 2 + ret i1 %cmp +} + +define i1 @range_check_to_icmp_eq2(i32 range(i32 -1, 2) %x) { +; CHECK-LABEL: define i1 @range_check_to_icmp_eq2( +; CHECK-SAME: i32 range(i32 -1, 2) [[X:%.*]]) { +; CHECK-NEXT: [[OFF:%.*]] = add nsw i32 [[X]], -1 +; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[X]], 1 +; CHECK-NEXT: ret i1 [[CMP]] +; + %off = add nsw i32 %x, -1 + %cmp = icmp ult i32 %off, -2 + ret i1 %cmp +} + declare void @use(i8) diff --git a/llvm/test/Transforms/SLPVectorizer/RISCV/strided-loads-with-external-indices.ll b/llvm/test/Transforms/SLPVectorizer/RISCV/strided-loads-with-external-indices.ll index 655db54..a079203 100644 --- a/llvm/test/Transforms/SLPVectorizer/RISCV/strided-loads-with-external-indices.ll +++ b/llvm/test/Transforms/SLPVectorizer/RISCV/strided-loads-with-external-indices.ll @@ -10,14 +10,10 @@ define void @test() { ; CHECK-NEXT: [[SUB4_I_I65_US:%.*]] = or i64 0, 1 ; CHECK-NEXT: br label [[BODY:%.*]] ; CHECK: body: -; CHECK-NEXT: [[ADD_I_I62_US:%.*]] = shl i64 0, 0 -; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x i64> <i64 poison, i64 1>, i64 [[ADD_I_I62_US]], i32 0 -; CHECK-NEXT: [[TMP1:%.*]] = or <2 x i64> zeroinitializer, [[TMP0]] -; CHECK-NEXT: [[TMP2:%.*]] = getelementptr [[CLASS_A:%.*]], <2 x ptr> zeroinitializer, <2 x i64> [[TMP1]] -; CHECK-NEXT: [[TMP3:%.*]] = call <2 x i32> @llvm.masked.gather.v2i32.v2p0(<2 x ptr> [[TMP2]], i32 4, <2 x i1> splat (i1 true), <2 x i32> poison) -; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i32> [[TMP3]], i32 0 -; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i32> [[TMP3]], i32 1 -; CHECK-NEXT: [[CMP_I_I_I_I67_US:%.*]] = icmp slt i32 [[TMP4]], [[TMP5]] +; CHECK-NEXT: [[TMP0:%.*]] = call <2 x i32> @llvm.masked.gather.v2i32.v2p0(<2 x ptr> getelementptr ([[CLASS_A:%.*]], <2 x ptr> zeroinitializer, <2 x i64> <i64 0, i64 1>), i32 4, <2 x i1> splat (i1 true), <2 x i32> poison) +; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x i32> [[TMP0]], i32 0 +; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x i32> [[TMP0]], i32 1 +; CHECK-NEXT: [[CMP_I_I_I_I67_US:%.*]] = icmp slt i32 [[TMP1]], [[TMP2]] ; CHECK-NEXT: [[SPEC_SELECT_I_I68_US:%.*]] = select i1 false, i64 [[SUB4_I_I65_US]], i64 0 ; CHECK-NEXT: br label [[BODY]] ; diff --git a/llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-different-bitwidth.ll b/llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-different-bitwidth.ll index 7758596..87f2cca 100644 --- a/llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-different-bitwidth.ll +++ b/llvm/test/Transforms/SLPVectorizer/X86/ext-used-scalar-different-bitwidth.ll @@ -8,8 +8,8 @@ define i32 @test() { ; CHECK-NEXT: [[ENTRY:.*:]] ; CHECK-NEXT: store i32 152, ptr @f, align 4 ; CHECK-NEXT: [[AGG_TMP_SROA_0_0_COPYLOAD_I:%.*]] = load i32, ptr @f, align 4 -; CHECK-NEXT: [[ADD_I_I:%.*]] = shl i32 [[AGG_TMP_SROA_0_0_COPYLOAD_I]], 24 -; CHECK-NEXT: [[TMP0:%.*]] = insertelement <8 x i32> <i32 poison, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080>, i32 [[ADD_I_I]], i32 0 +; CHECK-NEXT: [[TMP3:%.*]] = insertelement <8 x i32> <i32 poison, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080, i32 83886080>, i32 [[AGG_TMP_SROA_0_0_COPYLOAD_I]], i32 0 +; CHECK-NEXT: [[TMP0:%.*]] = shl <8 x i32> [[TMP3]], <i32 24, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0> ; CHECK-NEXT: [[TMP1:%.*]] = add <8 x i32> <i32 83886080, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>, [[TMP0]] ; CHECK-NEXT: [[TMP2:%.*]] = ashr <8 x i32> [[TMP1]], splat (i32 24) ; CHECK-NEXT: [[TMP5:%.*]] = and <8 x i32> [[TMP2]], <i32 66440127, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> diff --git a/llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll b/llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll index 75aec45..3e0a374 100644 --- a/llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll +++ b/llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll @@ -247,32 +247,12 @@ entry: } define void @shl0(ptr noalias %dst, ptr noalias %src) { -; NON-POW2-LABEL: @shl0( -; NON-POW2-NEXT: entry: -; NON-POW2-NEXT: [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 1 -; NON-POW2-NEXT: [[TMP0:%.*]] = load i32, ptr [[SRC]], align 4 -; NON-POW2-NEXT: [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 1 -; NON-POW2-NEXT: store i32 [[TMP0]], ptr [[DST]], align 4 -; NON-POW2-NEXT: [[TMP1:%.*]] = load <3 x i32>, ptr [[INCDEC_PTR]], align 4 -; NON-POW2-NEXT: [[TMP2:%.*]] = shl <3 x i32> [[TMP1]], <i32 1, i32 2, i32 3> -; NON-POW2-NEXT: store <3 x i32> [[TMP2]], ptr [[INCDEC_PTR1]], align 4 -; NON-POW2-NEXT: ret void -; -; POW2-ONLY-LABEL: @shl0( -; POW2-ONLY-NEXT: entry: -; POW2-ONLY-NEXT: [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 1 -; POW2-ONLY-NEXT: [[TMP0:%.*]] = load i32, ptr [[SRC]], align 4 -; POW2-ONLY-NEXT: [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 1 -; POW2-ONLY-NEXT: store i32 [[TMP0]], ptr [[DST]], align 4 -; POW2-ONLY-NEXT: [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 3 -; POW2-ONLY-NEXT: [[INCDEC_PTR6:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 3 -; POW2-ONLY-NEXT: [[TMP1:%.*]] = load <2 x i32>, ptr [[INCDEC_PTR]], align 4 -; POW2-ONLY-NEXT: [[TMP2:%.*]] = shl <2 x i32> [[TMP1]], <i32 1, i32 2> -; POW2-ONLY-NEXT: store <2 x i32> [[TMP2]], ptr [[INCDEC_PTR1]], align 4 -; POW2-ONLY-NEXT: [[TMP3:%.*]] = load i32, ptr [[INCDEC_PTR4]], align 4 -; POW2-ONLY-NEXT: [[SHL8:%.*]] = shl i32 [[TMP3]], 3 -; POW2-ONLY-NEXT: store i32 [[SHL8]], ptr [[INCDEC_PTR6]], align 4 -; POW2-ONLY-NEXT: ret void +; CHECK-LABEL: @shl0( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = shl <4 x i32> [[TMP0]], <i32 0, i32 1, i32 2, i32 3> +; CHECK-NEXT: store <4 x i32> [[TMP1]], ptr [[DST:%.*]], align 4 +; CHECK-NEXT: ret void ; entry: %incdec.ptr = getelementptr inbounds i32, ptr %src, i64 1 diff --git a/llvm/test/Transforms/SLPVectorizer/bool-logical-op-reduction-with-poison.ll b/llvm/test/Transforms/SLPVectorizer/bool-logical-op-reduction-with-poison.ll index a5b1e9b..769b360 100644 --- a/llvm/test/Transforms/SLPVectorizer/bool-logical-op-reduction-with-poison.ll +++ b/llvm/test/Transforms/SLPVectorizer/bool-logical-op-reduction-with-poison.ll @@ -1,25 +1,44 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3 -; RUN: %if x86-registered-target %{ opt -S --passes=slp-vectorizer < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s %} -; RUN: %if aarch64-registered-target %{ opt -S --passes=slp-vectorizer < %s -mtriple=aarch64-unknown-linux-gnu | FileCheck %s %} +; RUN: %if x86-registered-target %{ opt -S --passes=slp-vectorizer < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s --check-prefix=X86 %} +; RUN: %if aarch64-registered-target %{ opt -S --passes=slp-vectorizer < %s -mtriple=aarch64-unknown-linux-gnu | FileCheck %s --check-prefix=AARCH64 %} define i1 @test(i32 %0, i32 %1, i32 %p) { -; CHECK-LABEL: define i1 @test( -; CHECK-SAME: i32 [[TMP0:%.*]], i32 [[TMP1:%.*]], i32 [[P:%.*]]) { -; CHECK-NEXT: entry: -; CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[TMP0]], 0 -; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i32 0 -; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> zeroinitializer -; CHECK-NEXT: [[TMP4:%.*]] = shl <4 x i32> zeroinitializer, [[TMP3]] -; CHECK-NEXT: [[TMP5:%.*]] = icmp slt <4 x i32> [[TMP4]], zeroinitializer -; CHECK-NEXT: [[CMP6:%.*]] = icmp slt i32 0, [[P]] -; CHECK-NEXT: [[TMP6:%.*]] = freeze <4 x i1> [[TMP5]] -; CHECK-NEXT: [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]]) -; CHECK-NEXT: [[OP_RDX:%.*]] = select i1 [[TMP7]], i1 true, i1 [[CMP6]] -; CHECK-NEXT: [[OP_RDX1:%.*]] = select i1 [[CMP1]], i1 true, i1 [[CMP1]] -; CHECK-NEXT: [[TMP8:%.*]] = freeze i1 [[OP_RDX]] -; CHECK-NEXT: [[OP_RDX2:%.*]] = select i1 [[TMP8]], i1 true, i1 [[OP_RDX1]] -; CHECK-NEXT: ret i1 [[OP_RDX2]] +; X86-LABEL: define i1 @test( +; X86-SAME: i32 [[TMP0:%.*]], i32 [[TMP1:%.*]], i32 [[P:%.*]]) { +; X86-NEXT: entry: +; X86-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[TMP0]], 0 +; X86-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i32 0 +; X86-NEXT: [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> zeroinitializer +; X86-NEXT: [[TMP4:%.*]] = shl <4 x i32> zeroinitializer, [[TMP3]] +; X86-NEXT: [[TMP5:%.*]] = icmp slt <4 x i32> [[TMP4]], zeroinitializer +; X86-NEXT: [[CMP6:%.*]] = icmp slt i32 0, [[P]] +; X86-NEXT: [[TMP6:%.*]] = freeze <4 x i1> [[TMP5]] +; X86-NEXT: [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]]) +; X86-NEXT: [[OP_RDX:%.*]] = select i1 [[TMP7]], i1 true, i1 [[CMP6]] +; X86-NEXT: [[OP_RDX1:%.*]] = select i1 [[CMP1]], i1 true, i1 [[CMP1]] +; X86-NEXT: [[TMP8:%.*]] = freeze i1 [[OP_RDX]] +; X86-NEXT: [[OP_RDX2:%.*]] = select i1 [[TMP8]], i1 true, i1 [[OP_RDX1]] +; X86-NEXT: ret i1 [[OP_RDX2]] +; +; AARCH64-LABEL: define i1 @test( +; AARCH64-SAME: i32 [[TMP0:%.*]], i32 [[TMP1:%.*]], i32 [[P:%.*]]) { +; AARCH64-NEXT: entry: +; AARCH64-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[TMP0]], 0 +; AARCH64-NEXT: [[SHL4:%.*]] = shl i32 0, [[TMP1]] +; AARCH64-NEXT: [[CMP5:%.*]] = icmp slt i32 [[SHL4]], 0 +; AARCH64-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> <i32 0, i32 poison, i32 poison, i32 poison>, i32 [[TMP1]], i32 1 +; AARCH64-NEXT: [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 1> +; AARCH64-NEXT: [[TMP4:%.*]] = shl <4 x i32> zeroinitializer, [[TMP3]] +; AARCH64-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> <i32 poison, i32 0, i32 0, i32 0>, i32 [[P]], i32 0 +; AARCH64-NEXT: [[TMP6:%.*]] = icmp slt <4 x i32> [[TMP4]], [[TMP5]] +; AARCH64-NEXT: [[TMP7:%.*]] = freeze <4 x i1> [[TMP6]] +; AARCH64-NEXT: [[TMP8:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP7]]) +; AARCH64-NEXT: [[OP_RDX:%.*]] = select i1 [[TMP8]], i1 true, i1 [[CMP5]] +; AARCH64-NEXT: [[OP_RDX1:%.*]] = select i1 [[CMP1]], i1 true, i1 [[CMP1]] +; AARCH64-NEXT: [[TMP9:%.*]] = freeze i1 [[OP_RDX]] +; AARCH64-NEXT: [[OP_RDX2:%.*]] = select i1 [[TMP9]], i1 true, i1 [[OP_RDX1]] +; AARCH64-NEXT: ret i1 [[OP_RDX2]] ; entry: %cmp1 = icmp sgt i32 %0, 0 diff --git a/llvm/test/Transforms/SimplifyCFG/indirectbr.ll b/llvm/test/Transforms/SimplifyCFG/indirectbr.ll index 87d8b39..2fa36b0 100644 --- a/llvm/test/Transforms/SimplifyCFG/indirectbr.ll +++ b/llvm/test/Transforms/SimplifyCFG/indirectbr.ll @@ -1,4 +1,4 @@ -; NOTE: Assertions have been autogenerated by utils/update_test_checks.py +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals ; RUN: opt -S -passes=simplifycfg -simplifycfg-require-and-preserve-domtree=1 < %s | FileCheck %s ; SimplifyCFG should eliminate redundant indirectbr edges. @@ -8,7 +8,11 @@ declare void @A() declare void @B(i32) declare void @C() -define void @indbrtest0(ptr %P, ptr %Q) { +;. +; CHECK: @anchor = constant [13 x ptr] [ptr blockaddress(@indbrtest3, %L1), ptr blockaddress(@indbrtest3, %L2), ptr inttoptr (i32 1 to ptr), ptr blockaddress(@indbrtest4, %L1), ptr inttoptr (i32 1 to ptr), ptr inttoptr (i32 1 to ptr), ptr inttoptr (i32 1 to ptr), ptr inttoptr (i32 1 to ptr), ptr inttoptr (i32 1 to ptr), ptr inttoptr (i32 1 to ptr), ptr inttoptr (i32 1 to ptr), ptr inttoptr (i32 1 to ptr), ptr inttoptr (i32 1 to ptr)] +; CHECK: @xblkx.bbs = internal unnamed_addr constant [9 x ptr] [ptr blockaddress(@indbrtest7, %xlab4x), ptr blockaddress(@indbrtest7, %xlab4x), ptr blockaddress(@indbrtest7, %v2j), ptr blockaddress(@indbrtest7, %xlab4x), ptr blockaddress(@indbrtest7, %xlab4x), ptr blockaddress(@indbrtest7, %xlab4x), ptr blockaddress(@indbrtest7, %xlab4x), ptr blockaddress(@indbrtest7, %xlab4x), ptr blockaddress(@indbrtest7, %v2j)] +;. +define void @indbrtest0(ptr %P, ptr %Q) !prof !0 { ; CHECK-LABEL: @indbrtest0( ; CHECK-NEXT: entry: ; CHECK-NEXT: store ptr blockaddress(@indbrtest0, [[BB0:%.*]]), ptr [[P:%.*]], align 8 @@ -16,7 +20,7 @@ define void @indbrtest0(ptr %P, ptr %Q) { ; CHECK-NEXT: store ptr blockaddress(@indbrtest0, [[BB2:%.*]]), ptr [[P]], align 8 ; CHECK-NEXT: call void @foo() ; CHECK-NEXT: [[T:%.*]] = load ptr, ptr [[Q:%.*]], align 8 -; CHECK-NEXT: indirectbr ptr [[T]], [label [[BB0]], label [[BB1]], label %BB2] +; CHECK-NEXT: indirectbr ptr [[T]], [label [[BB0]], label [[BB1]], label %BB2], !prof [[PROF1:![0-9]+]] ; CHECK: BB0: ; CHECK-NEXT: call void @A() ; CHECK-NEXT: br label [[BB1]] @@ -36,7 +40,7 @@ entry: store ptr blockaddress(@indbrtest0, %BB2), ptr %P call void @foo() %t = load ptr, ptr %Q - indirectbr ptr %t, [label %BB0, label %BB1, label %BB2, label %BB0, label %BB1, label %BB2] + indirectbr ptr %t, [label %BB0, label %BB1, label %BB2, label %BB0, label %BB1, label %BB2], !prof !1 BB0: call void @A() br label %BB1 @@ -103,10 +107,10 @@ BB0: ; SimplifyCFG should turn the indirectbr into a conditional branch on the ; condition of the select. -define void @indbrtest3(i1 %cond, ptr %address) nounwind { +define void @indbrtest3(i1 %cond, ptr %address) nounwind !prof !0 { ; CHECK-LABEL: @indbrtest3( ; CHECK-NEXT: entry: -; CHECK-NEXT: br i1 [[COND:%.*]], label [[L1:%.*]], label [[L2:%.*]] +; CHECK-NEXT: br i1 [[COND:%.*]], label [[L1:%.*]], label [[L2:%.*]], !prof [[PROF2:![0-9]+]] ; CHECK: common.ret: ; CHECK-NEXT: ret void ; CHECK: L1: @@ -117,8 +121,8 @@ define void @indbrtest3(i1 %cond, ptr %address) nounwind { ; CHECK-NEXT: br label [[COMMON_RET]] ; entry: - %indirect.goto.dest = select i1 %cond, ptr blockaddress(@indbrtest3, %L1), ptr blockaddress(@indbrtest3, %L2) - indirectbr ptr %indirect.goto.dest, [label %L1, label %L2, label %L3] + %indirect.goto.dest = select i1 %cond, ptr blockaddress(@indbrtest3, %L1), ptr blockaddress(@indbrtest3, %L2), !prof !2 + indirectbr ptr %indirect.goto.dest, [label %L1, label %L2, label %L3], !prof !3 L1: call void @A() @@ -385,3 +389,15 @@ declare i32 @xfunc5x() declare i8 @xfunc7x() declare i32 @xselectorx() declare i32 @xactionx() + +!0 = !{!"function_entry_count", i32 10} +!1 = !{!"branch_weights", i32 3, i32 5, i32 7, i32 11, i32 13, i32 17} +!2 = !{!"branch_weights", i32 3, i32 5} +!3 = !{!"branch_weights", i32 3, i32 5, i32 7} +;. +; CHECK: attributes #[[ATTR0:[0-9]+]] = { nounwind } +;. +; CHECK: [[META0:![0-9]+]] = !{!"function_entry_count", i32 10} +; CHECK: [[PROF1]] = !{!"branch_weights", i32 14, i32 18, i32 24} +; CHECK: [[PROF2]] = !{!"branch_weights", i32 3, i32 5} +;. diff --git a/llvm/test/Transforms/SimplifyCFG/switch-dead-default.ll b/llvm/test/Transforms/SimplifyCFG/switch-dead-default.ll index 4a457cc..a0e29dd 100644 --- a/llvm/test/Transforms/SimplifyCFG/switch-dead-default.ll +++ b/llvm/test/Transforms/SimplifyCFG/switch-dead-default.ll @@ -7,8 +7,7 @@ declare void @foo(i32) define void @test(i1 %a) { ; CHECK-LABEL: define void @test( ; CHECK-SAME: i1 [[A:%.*]]) { -; CHECK-NEXT: [[A_OFF:%.*]] = add i1 [[A]], true -; CHECK-NEXT: [[SWITCH:%.*]] = icmp ult i1 [[A_OFF]], true +; CHECK-NEXT: [[SWITCH:%.*]] = icmp eq i1 [[A]], true ; CHECK-NEXT: br i1 [[SWITCH]], label [[TRUE:%.*]], label [[FALSE:%.*]] ; CHECK: common.ret: ; CHECK-NEXT: ret void @@ -209,8 +208,7 @@ define void @test5(i8 %a) { ; CHECK-SAME: i8 [[A:%.*]]) { ; CHECK-NEXT: [[CMP:%.*]] = icmp ult i8 [[A]], 2 ; CHECK-NEXT: call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: [[A_OFF:%.*]] = add i8 [[A]], -1 -; CHECK-NEXT: [[SWITCH:%.*]] = icmp ult i8 [[A_OFF]], 1 +; CHECK-NEXT: [[SWITCH:%.*]] = icmp eq i8 [[A]], 1 ; CHECK-NEXT: br i1 [[SWITCH]], label [[TRUE:%.*]], label [[FALSE:%.*]] ; CHECK: common.ret: ; CHECK-NEXT: ret void @@ -243,8 +241,7 @@ define void @test6(i8 %a) { ; CHECK-NEXT: [[AND:%.*]] = and i8 [[A]], -2 ; CHECK-NEXT: [[CMP:%.*]] = icmp eq i8 [[AND]], -2 ; CHECK-NEXT: call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: [[A_OFF:%.*]] = add i8 [[A]], 1 -; CHECK-NEXT: [[SWITCH:%.*]] = icmp ult i8 [[A_OFF]], 1 +; CHECK-NEXT: [[SWITCH:%.*]] = icmp eq i8 [[A]], -1 ; CHECK-NEXT: br i1 [[SWITCH]], label [[TRUE:%.*]], label [[FALSE:%.*]] ; CHECK: common.ret: ; CHECK-NEXT: ret void @@ -279,8 +276,7 @@ define void @test7(i8 %a) { ; CHECK-NEXT: [[AND:%.*]] = and i8 [[A]], -2 ; CHECK-NEXT: [[CMP:%.*]] = icmp eq i8 [[AND]], -2 ; CHECK-NEXT: call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: [[A_OFF:%.*]] = add i8 [[A]], 1 -; CHECK-NEXT: [[SWITCH:%.*]] = icmp ult i8 [[A_OFF]], 1 +; CHECK-NEXT: [[SWITCH:%.*]] = icmp eq i8 [[A]], -1 ; CHECK-NEXT: br i1 [[SWITCH]], label [[TRUE:%.*]], label [[FALSE:%.*]] ; CHECK: common.ret: ; CHECK-NEXT: ret void diff --git a/llvm/test/Transforms/SimplifyCFG/switch-range-to-icmp.ll b/llvm/test/Transforms/SimplifyCFG/switch-range-to-icmp.ll index 8f2ae2d..0fc3c19 100644 --- a/llvm/test/Transforms/SimplifyCFG/switch-range-to-icmp.ll +++ b/llvm/test/Transforms/SimplifyCFG/switch-range-to-icmp.ll @@ -188,4 +188,217 @@ exit: ret void } +define i32 @wrapping_known_range(i8 range(i8 0, 6) %arg) { +; CHECK-LABEL: @wrapping_known_range( +; CHECK-NEXT: [[ARG_OFF:%.*]] = add i8 [[ARG:%.*]], -1 +; CHECK-NEXT: [[SWITCH:%.*]] = icmp ult i8 [[ARG_OFF]], 3 +; CHECK-NEXT: br i1 [[SWITCH]], label [[ELSE:%.*]], label [[IF:%.*]] +; CHECK: common.ret: +; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ [[I0:%.*]], [[IF]] ], [ [[I1:%.*]], [[ELSE]] ] +; CHECK-NEXT: ret i32 [[COMMON_RET_OP]] +; CHECK: if: +; CHECK-NEXT: [[I0]] = call i32 @f(i32 0) +; CHECK-NEXT: br label [[COMMON_RET:%.*]] +; CHECK: else: +; CHECK-NEXT: [[I1]] = call i32 @f(i32 1) +; CHECK-NEXT: br label [[COMMON_RET]] +; + switch i8 %arg, label %else [ + i8 0, label %if + i8 4, label %if + i8 5, label %if + ] + +if: + %i0 = call i32 @f(i32 0) + ret i32 %i0 + +else: + %i1 = call i32 @f(i32 1) + ret i32 %i1 +} + +define i32 @wrapping_known_range_2(i8 range(i8 0, 6) %arg) { +; CHECK-LABEL: @wrapping_known_range_2( +; CHECK-NEXT: [[SWITCH:%.*]] = icmp eq i8 [[ARG:%.*]], 1 +; CHECK-NEXT: br i1 [[SWITCH]], label [[ELSE:%.*]], label [[IF:%.*]] +; CHECK: common.ret: +; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ [[I0:%.*]], [[IF]] ], [ [[I1:%.*]], [[ELSE]] ] +; CHECK-NEXT: ret i32 [[COMMON_RET_OP]] +; CHECK: if: +; CHECK-NEXT: [[I0]] = call i32 @f(i32 0) +; CHECK-NEXT: br label [[COMMON_RET:%.*]] +; CHECK: else: +; CHECK-NEXT: [[I1]] = call i32 @f(i32 1) +; CHECK-NEXT: br label [[COMMON_RET]] +; + switch i8 %arg, label %else [ + i8 0, label %if + i8 2, label %if + i8 3, label %if + i8 4, label %if + i8 5, label %if + ] + +if: + %i0 = call i32 @f(i32 0) + ret i32 %i0 + +else: + %i1 = call i32 @f(i32 1) + ret i32 %i1 +} + +define i32 @wrapping_range(i8 %arg) { +; CHECK-LABEL: @wrapping_range( +; CHECK-NEXT: [[ARG_OFF:%.*]] = add i8 [[ARG:%.*]], -1 +; CHECK-NEXT: [[SWITCH:%.*]] = icmp ult i8 [[ARG_OFF]], -4 +; CHECK-NEXT: br i1 [[SWITCH]], label [[ELSE:%.*]], label [[IF:%.*]] +; CHECK: common.ret: +; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ [[I0:%.*]], [[IF]] ], [ [[I1:%.*]], [[ELSE]] ] +; CHECK-NEXT: ret i32 [[COMMON_RET_OP]] +; CHECK: if: +; CHECK-NEXT: [[I0]] = call i32 @f(i32 0) +; CHECK-NEXT: br label [[COMMON_RET:%.*]] +; CHECK: else: +; CHECK-NEXT: [[I1]] = call i32 @f(i32 1) +; CHECK-NEXT: br label [[COMMON_RET]] +; + switch i8 %arg, label %else [ + i8 0, label %if + i8 -3, label %if + i8 -2, label %if + i8 -1, label %if + ] + +if: + %i0 = call i32 @f(i32 0) + ret i32 %i0 + +else: + %i1 = call i32 @f(i32 1) + ret i32 %i1 +} + +define i8 @wrapping_range_phi(i8 %arg) { +; CHECK-LABEL: @wrapping_range_phi( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[ARG_OFF:%.*]] = add i8 [[ARG:%.*]], -1 +; CHECK-NEXT: [[SWITCH:%.*]] = icmp ult i8 [[ARG_OFF]], -2 +; CHECK-NEXT: [[SPEC_SELECT:%.*]] = select i1 [[SWITCH]], i8 0, i8 1 +; CHECK-NEXT: ret i8 [[SPEC_SELECT]] +; +entry: + switch i8 %arg, label %else [ + i8 0, label %if + i8 -1, label %if + ] + +if: + %i = phi i8 [ 0, %else ], [ 1, %entry ], [ 1, %entry ] + ret i8 %i + +else: + br label %if +} + +define i32 @no_continuous_wrapping_range(i8 %arg) { +; CHECK-LABEL: @no_continuous_wrapping_range( +; CHECK-NEXT: switch i8 [[ARG:%.*]], label [[ELSE:%.*]] [ +; CHECK-NEXT: i8 0, label [[IF:%.*]] +; CHECK-NEXT: i8 -3, label [[IF]] +; CHECK-NEXT: i8 -1, label [[IF]] +; CHECK-NEXT: ] +; CHECK: common.ret: +; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ [[I0:%.*]], [[IF]] ], [ [[I1:%.*]], [[ELSE]] ] +; CHECK-NEXT: ret i32 [[COMMON_RET_OP]] +; CHECK: if: +; CHECK-NEXT: [[I0]] = call i32 @f(i32 0) +; CHECK-NEXT: br label [[COMMON_RET:%.*]] +; CHECK: else: +; CHECK-NEXT: [[I1]] = call i32 @f(i32 1) +; CHECK-NEXT: br label [[COMMON_RET]] +; + switch i8 %arg, label %else [ + i8 0, label %if + i8 -3, label %if + i8 -1, label %if + ] + +if: + %i0 = call i32 @f(i32 0) + ret i32 %i0 + +else: + %i1 = call i32 @f(i32 1) + ret i32 %i1 +} + +define i32 @one_case_1(i32 %x) { +; CHECK-LABEL: @one_case_1( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[SWITCH:%.*]] = icmp eq i32 [[X:%.*]], 10 +; CHECK-NEXT: br i1 [[SWITCH]], label [[A:%.*]], label [[B:%.*]] +; CHECK: common.ret: +; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ [[TMP0:%.*]], [[B]] ], [ [[TMP1:%.*]], [[A]] ] +; CHECK-NEXT: ret i32 [[COMMON_RET_OP]] +; CHECK: a: +; CHECK-NEXT: [[TMP0]] = call i32 @f(i32 0) +; CHECK-NEXT: br label [[COMMON_RET:%.*]] +; CHECK: b: +; CHECK-NEXT: [[TMP1]] = call i32 @f(i32 1) +; CHECK-NEXT: br label [[COMMON_RET]] +; +entry: + switch i32 %x, label %unreachable [ + i32 5, label %a + i32 6, label %a + i32 7, label %a + i32 10, label %b + ] + +unreachable: + unreachable +a: + %0 = call i32 @f(i32 0) + ret i32 %0 +b: + %1 = call i32 @f(i32 1) + ret i32 %1 +} + +define i32 @one_case_2(i32 %x) { +; CHECK-LABEL: @one_case_2( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[SWITCH:%.*]] = icmp eq i32 [[X:%.*]], 5 +; CHECK-NEXT: br i1 [[SWITCH]], label [[A:%.*]], label [[B:%.*]] +; CHECK: common.ret: +; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ [[TMP0:%.*]], [[A]] ], [ [[TMP1:%.*]], [[B]] ] +; CHECK-NEXT: ret i32 [[COMMON_RET_OP]] +; CHECK: a: +; CHECK-NEXT: [[TMP0]] = call i32 @f(i32 0) +; CHECK-NEXT: br label [[COMMON_RET:%.*]] +; CHECK: b: +; CHECK-NEXT: [[TMP1]] = call i32 @f(i32 1) +; CHECK-NEXT: br label [[COMMON_RET]] +; +entry: + switch i32 %x, label %unreachable [ + i32 5, label %a + i32 10, label %b + i32 11, label %b + i32 12, label %b + i32 13, label %b + ] + +unreachable: + unreachable +a: + %0 = call i32 @f(i32 0) + ret i32 %0 +b: + %1 = call i32 @f(i32 1) + ret i32 %1 +} + declare void @bar(ptr nonnull dereferenceable(4)) |