diff options
| -rw-r--r-- | llvm/include/llvm/IR/InstrTypes.h | 9 | ||||
| -rw-r--r-- | llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp | 6 | ||||
| -rw-r--r-- | llvm/test/Transforms/InstCombine/fast-math-scalarization.ll | 39 |
3 files changed, 52 insertions, 2 deletions
diff --git a/llvm/include/llvm/IR/InstrTypes.h b/llvm/include/llvm/IR/InstrTypes.h index 5091bb40783..ef86eba8135 100644 --- a/llvm/include/llvm/IR/InstrTypes.h +++ b/llvm/include/llvm/IR/InstrTypes.h @@ -386,6 +386,15 @@ public: } #include "llvm/IR/Instruction.def" + static BinaryOperator *CreateWithCopiedFlags(BinaryOps Opc, + Value *V1, Value *V2, + BinaryOperator *CopyBO, + const Twine &Name = "") { + BinaryOperator *BO = Create(Opc, V1, V2, Name); + BO->copyIRFlags(CopyBO); + return BO; + } + static BinaryOperator *CreateNSW(BinaryOps Opc, Value *V1, Value *V2, const Twine &Name = "") { BinaryOperator *BO = Create(Opc, V1, V2, Name); diff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp index 258d7b5762d..d92806f08de 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp @@ -106,7 +106,8 @@ Instruction *InstCombiner::scalarizePHI(ExtractElementInst &EI, PHINode *PN) { B0->getOperand(opId)->getName() + ".Elt"), *B0); Value *newPHIUser = InsertNewInstWith( - BinaryOperator::Create(B0->getOpcode(), scalarPHI, Op), *B0); + BinaryOperator::CreateWithCopiedFlags(B0->getOpcode(), + scalarPHI, Op, B0), *B0); scalarPHI->addIncoming(newPHIUser, inBB); } else { // Scalarize PHI input: @@ -193,7 +194,8 @@ Instruction *InstCombiner::visitExtractElementInst(ExtractElementInst &EI) { Value *newEI1 = Builder->CreateExtractElement(BO->getOperand(1), EI.getOperand(1), EI.getName()+".rhs"); - return BinaryOperator::Create(BO->getOpcode(), newEI0, newEI1); + return BinaryOperator::CreateWithCopiedFlags(BO->getOpcode(), + newEI0, newEI1, BO); } } else if (InsertElementInst *IE = dyn_cast<InsertElementInst>(I)) { // Extracting the inserted element? diff --git a/llvm/test/Transforms/InstCombine/fast-math-scalarization.ll b/llvm/test/Transforms/InstCombine/fast-math-scalarization.ll new file mode 100644 index 00000000000..406ebebfdd2 --- /dev/null +++ b/llvm/test/Transforms/InstCombine/fast-math-scalarization.ll @@ -0,0 +1,39 @@ +; RUN: opt -instcombine -S < %s | FileCheck %s + +; CHECK-LABEL: test_scalarize_phi +; CHECK: fmul fast float +define void @test_scalarize_phi(i32 * %n, float * %inout) { +entry: + %0 = load volatile float, float * %inout, align 4 + %splat.splatinsert = insertelement <4 x float> undef, float %0, i32 0 + %splat.splat = shufflevector <4 x float> %splat.splatinsert, <4 x float> undef, <4 x i32> zeroinitializer + %splat.splatinsert1 = insertelement <4 x float> undef, float 3.0, i32 0 + br label %for.cond + +for.cond: + %x.0 = phi <4 x float> [ %splat.splat, %entry ], [ %mul, %for.body ] + %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ] + %1 = load i32, i32 * %n, align 4 + %cmp = icmp ne i32 %i.0, %1 + br i1 %cmp, label %for.body, label %for.end + +for.body: + %2 = extractelement <4 x float> %x.0, i32 1 + store volatile float %2, float * %inout, align 4 + %mul = fmul fast <4 x float> %x.0, <float 0x4002A3D700000000, float 0x4002A3D700000000, float 0x4002A3D700000000, float 0x4002A3D700000000> + %inc = add nsw i32 %i.0, 1 + br label %for.cond + +for.end: + ret void +} + +; CHECK-LABEL: test_extract_element_fastmath +; CHECK: fadd fast float +define float @test_extract_element_fastmath(<4 x float> %x) #0 { +entry: + %add = fadd fast <4 x float> %x, <float 0x4002A3D700000000, float 0x4002A3D700000000, float 0x4002A3D700000000, float 0x4002A3D700000000> + %0 = extractelement <4 x float> %add, i32 2 + ret float %0 +} + |
