[llvm] [DirectX] legalize memset (PR #136244)
Farzon Lotfi via llvm-commits
llvm-commits at lists.llvm.org
Mon Apr 21 15:29:37 PDT 2025
https://github.com/farzonl updated https://github.com/llvm/llvm-project/pull/136244
>From cc23d39801e82c47b232063a0883d664ea85ac08 Mon Sep 17 00:00:00 2001
From: Farzon Lotfi <farzonlotfi at microsoft.com>
Date: Thu, 17 Apr 2025 22:46:29 -0400
Subject: [PATCH 1/2] [DirectX] legalize memset fixes #136243
This change converts memset into a series of geps and stores
It is intentionally limited to memsets of fixed size
It also converts the byte stores to type stores.
DXIL does not support i8 plus this reduces the total number of gep and
store instructions.
This change also moves DXILFinalizeLinkage to run after Legalization to
clean up any dead intrinsic definitions.
---
llvm/lib/Target/DirectX/DXILLegalizePass.cpp | 78 ++++++++++++++++++
.../Target/DirectX/DirectXTargetMachine.cpp | 2 +-
llvm/test/CodeGen/DirectX/legalize-memset.ll | 81 +++++++++++++++++++
llvm/test/CodeGen/DirectX/llc-pipeline.ll | 2 +-
4 files changed, 161 insertions(+), 2 deletions(-)
create mode 100644 llvm/test/CodeGen/DirectX/legalize-memset.ll
diff --git a/llvm/lib/Target/DirectX/DXILLegalizePass.cpp b/llvm/lib/Target/DirectX/DXILLegalizePass.cpp
index b62ff4c52f70c..d55a84d0126ec 100644
--- a/llvm/lib/Target/DirectX/DXILLegalizePass.cpp
+++ b/llvm/lib/Target/DirectX/DXILLegalizePass.cpp
@@ -12,6 +12,7 @@
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/InstIterator.h"
#include "llvm/IR/Instruction.h"
+#include "llvm/IR/Module.h"
#include "llvm/Pass.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include <functional>
@@ -151,6 +152,82 @@ downcastI64toI32InsertExtractElements(Instruction &I,
}
}
+void emitMemset(IRBuilder<> &Builder, Value *Dst, Value *Val,
+ ConstantInt *SizeCI) {
+ LLVMContext &Ctx = Builder.getContext();
+ [[maybe_unused]] DataLayout DL =
+ Builder.GetInsertBlock()->getModule()->getDataLayout();
+ [[maybe_unused]] uint64_t OrigSize = SizeCI->getZExtValue();
+
+ AllocaInst *Alloca = dyn_cast<AllocaInst>(Dst);
+
+ assert(Alloca && "Expected memset on an Alloca");
+ assert(OrigSize == Alloca->getAllocationSize(DL)->getFixedValue() &&
+ "Expected for memset size to match DataLayout size");
+
+ Type *AllocatedTy = Alloca->getAllocatedType();
+ ArrayType *ArrTy = dyn_cast<ArrayType>(AllocatedTy);
+ assert(ArrTy && "Expected Alloca for an Array Type");
+
+ Type *ElemTy = ArrTy->getElementType();
+ uint64_t Size = ArrTy->getArrayNumElements();
+
+ [[maybe_unused]] uint64_t ElemSize = DL.getTypeStoreSize(ElemTy);
+
+ assert(ElemSize > 0 && "Size must be set");
+ assert(OrigSize == ElemSize * Size && "Size in bytes must match");
+
+ Value *TypedVal = Val;
+ if (Val->getType() != ElemTy)
+ TypedVal = Builder.CreateIntCast(Val, ElemTy,
+ false); // Or use CreateBitCast for float
+
+ for (uint64_t I = 0; I < Size; ++I) {
+ Value *Offset = ConstantInt::get(Type::getInt32Ty(Ctx), I);
+ Value *Ptr = Builder.CreateGEP(ElemTy, Dst, Offset, "gep");
+ Builder.CreateStore(TypedVal, Ptr);
+ }
+}
+
+void removeLifetimesForMemset(CallInst *Memset,
+ SmallVectorImpl<Instruction *> &ToRemove) {
+ assert(Memset->getCalledFunction()->getIntrinsicID() == Intrinsic::memset &&
+ "Expected a memset intrinsic");
+
+ Value *DstPtr = Memset->getArgOperand(0);
+ DstPtr = DstPtr->stripPointerCasts();
+
+ for (User *U : DstPtr->users()) {
+ if (auto *CI = dyn_cast<CallInst>(U)) {
+ switch (CI->getIntrinsicID()) {
+ case Intrinsic::lifetime_start:
+ case Intrinsic::lifetime_end:
+ ToRemove.push_back(CI);
+ break;
+ }
+ }
+ }
+}
+
+static void removeMemSet(Instruction &I,
+ SmallVectorImpl<Instruction *> &ToRemove,
+ DenseMap<Value *, Value *>) {
+ if (CallInst *CI = dyn_cast<CallInst>(&I)) {
+ Intrinsic::ID ID = CI->getIntrinsicID();
+ if (ID == Intrinsic::memset) {
+ IRBuilder<> Builder(&I);
+ Value *Dst = CI->getArgOperand(0);
+ Value *Val = CI->getArgOperand(1);
+ [[maybe_unused]] ConstantInt *Size =
+ dyn_cast<ConstantInt>(CI->getArgOperand(2));
+ assert(Size && "Expected Size to be a ConstantInt");
+ emitMemset(Builder, Dst, Val, Size);
+ removeLifetimesForMemset(CI, ToRemove);
+ ToRemove.push_back(CI);
+ }
+ }
+}
+
namespace {
class DXILLegalizationPipeline {
@@ -181,6 +258,7 @@ class DXILLegalizationPipeline {
LegalizationPipeline.push_back(fixI8TruncUseChain);
LegalizationPipeline.push_back(downcastI64toI32InsertExtractElements);
LegalizationPipeline.push_back(legalizeFreeze);
+ LegalizationPipeline.push_back(removeMemSet);
}
};
diff --git a/llvm/lib/Target/DirectX/DirectXTargetMachine.cpp b/llvm/lib/Target/DirectX/DirectXTargetMachine.cpp
index 41f6f37a41f9d..1cd21ae0fd55a 100644
--- a/llvm/lib/Target/DirectX/DirectXTargetMachine.cpp
+++ b/llvm/lib/Target/DirectX/DirectXTargetMachine.cpp
@@ -96,7 +96,6 @@ class DirectXPassConfig : public TargetPassConfig {
FunctionPass *createTargetRegisterAllocator(bool) override { return nullptr; }
void addCodeGenPrepare() override {
- addPass(createDXILFinalizeLinkageLegacyPass());
addPass(createDXILIntrinsicExpansionLegacyPass());
addPass(createDXILCBufferAccessLegacyPass());
addPass(createDXILDataScalarizationLegacyPass());
@@ -106,6 +105,7 @@ class DirectXPassConfig : public TargetPassConfig {
DxilScalarOptions.ScalarizeLoadStore = true;
addPass(createScalarizerPass(DxilScalarOptions));
addPass(createDXILLegalizeLegacyPass());
+ addPass(createDXILFinalizeLinkageLegacyPass());
addPass(createDXILTranslateMetadataLegacyPass());
addPass(createDXILOpLoweringLegacyPass());
addPass(createDXILPrepareModulePass());
diff --git a/llvm/test/CodeGen/DirectX/legalize-memset.ll b/llvm/test/CodeGen/DirectX/legalize-memset.ll
new file mode 100644
index 0000000000000..2d9b28753ef3d
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/legalize-memset.ll
@@ -0,0 +1,81 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -passes='dxil-legalize' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
+
+
+define void @replace_float_memset_test() {
+; CHECK-LABEL: define void @replace_float_memset_test() {
+; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [2 x float], align 4
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr float, ptr [[ACCUM_I_FLAT]], i32 0
+; CHECK-NEXT: store float 0.000000e+00, ptr [[GEP]], align 4
+; CHECK-NEXT: [[GEP1:%.*]] = getelementptr float, ptr [[ACCUM_I_FLAT]], i32 1
+; CHECK-NEXT: store float 0.000000e+00, ptr [[GEP1]], align 4
+; CHECK-NEXT: ret void
+;
+ %accum.i.flat = alloca [2 x float], align 4
+ call void @llvm.lifetime.start.p0(i64 8, ptr nonnull %accum.i.flat)
+ call void @llvm.memset.p0.i32(ptr nonnull align 4 dereferenceable(8) %accum.i.flat, i8 0, i32 8, i1 false)
+ call void @llvm.lifetime.end.p0(i64 8, ptr nonnull %accum.i.flat)
+ ret void
+}
+
+define void @replace_half_memset_test() {
+; CHECK-LABEL: define void @replace_half_memset_test() {
+; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [2 x half], align 4
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr half, ptr [[ACCUM_I_FLAT]], i32 0
+; CHECK-NEXT: store half 0xH0000, ptr [[GEP]], align 2
+; CHECK-NEXT: [[GEP1:%.*]] = getelementptr half, ptr [[ACCUM_I_FLAT]], i32 1
+; CHECK-NEXT: store half 0xH0000, ptr [[GEP1]], align 2
+; CHECK-NEXT: ret void
+;
+ %accum.i.flat = alloca [2 x half], align 4
+ call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %accum.i.flat)
+ call void @llvm.memset.p0.i32(ptr nonnull align 4 dereferenceable(8) %accum.i.flat, i8 0, i32 4, i1 false)
+ call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %accum.i.flat)
+ ret void
+}
+
+define void @replace_double_memset_test() {
+; CHECK-LABEL: define void @replace_double_memset_test() {
+; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [2 x double], align 4
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr double, ptr [[ACCUM_I_FLAT]], i32 0
+; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP]], align 8
+; CHECK-NEXT: [[GEP1:%.*]] = getelementptr double, ptr [[ACCUM_I_FLAT]], i32 1
+; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP1]], align 8
+; CHECK-NEXT: ret void
+;
+ %accum.i.flat = alloca [2 x double], align 4
+ call void @llvm.lifetime.start.p0(i64 16, ptr nonnull %accum.i.flat)
+ call void @llvm.memset.p0.i32(ptr nonnull align 4 dereferenceable(8) %accum.i.flat, i8 0, i32 16, i1 false)
+ call void @llvm.lifetime.end.p0(i64 16, ptr nonnull %accum.i.flat)
+ ret void
+}
+
+define void @replace_int16_memset_test() {
+; CHECK-LABEL: define void @replace_int16_memset_test() {
+; CHECK-NEXT: [[CACHE_I:%.*]] = alloca [2 x i16], align 2
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr i16, ptr [[CACHE_I]], i32 0
+; CHECK-NEXT: store i16 0, ptr [[GEP]], align 2
+; CHECK-NEXT: [[GEP1:%.*]] = getelementptr i16, ptr [[CACHE_I]], i32 1
+; CHECK-NEXT: store i16 0, ptr [[GEP1]], align 2
+; CHECK-NEXT: ret void
+;
+ %cache.i = alloca [2 x i16], align 2
+ call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %cache.i)
+ call void @llvm.memset.p0.i32(ptr nonnull align 2 dereferenceable(4) %cache.i, i8 0, i32 4, i1 false)
+ call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %cache.i)
+ ret void
+}
+
+define void @replace_int_memset_test() {
+; CHECK-LABEL: define void @replace_int_memset_test() {
+; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [1 x i32], align 4
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr i32, ptr [[ACCUM_I_FLAT]], i32 0
+; CHECK-NEXT: store i32 0, ptr [[GEP]], align 4
+; CHECK-NEXT: ret void
+;
+ %accum.i.flat = alloca [1 x i32], align 4
+ call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %accum.i.flat)
+ call void @llvm.memset.p0.i32(ptr nonnull align 4 dereferenceable(8) %accum.i.flat, i8 0, i32 4, i1 false)
+ call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %accum.i.flat)
+ ret void
+}
diff --git a/llvm/test/CodeGen/DirectX/llc-pipeline.ll b/llvm/test/CodeGen/DirectX/llc-pipeline.ll
index b1bd9f16f4efa..6b4a39a6d49ad 100644
--- a/llvm/test/CodeGen/DirectX/llc-pipeline.ll
+++ b/llvm/test/CodeGen/DirectX/llc-pipeline.ll
@@ -13,7 +13,6 @@
; CHECK-OBJ-NEXT: Create Garbage Collector Module Metadata
; CHECK-NEXT: ModulePass Manager
-; CHECK-NEXT: DXIL Finalize Linkage
; CHECK-NEXT: DXIL Intrinsic Expansion
; CHECK-NEXT: DXIL CBuffer Access
; CHECK-NEXT: DXIL Data Scalarization
@@ -23,6 +22,7 @@
; CHECK-NEXT: Dominator Tree Construction
; CHECK-NEXT: Scalarize vector operations
; CHECK-NEXT: DXIL Legalizer
+; CHECK-NEXT: DXIL Finalize Linkage
; CHECK-NEXT: DXIL Resource Binding Analysis
; CHECK-NEXT: DXIL Module Metadata analysis
; CHECK-NEXT: DXIL Shader Flag Analysis
>From 9f65060a7125345da0363c068d46a9b23b00a1be Mon Sep 17 00:00:00 2001
From: Farzon Lotfi <farzonlotfi at microsoft.com>
Date: Mon, 21 Apr 2025 18:28:17 -0400
Subject: [PATCH 2/2] remove lifetime intrinsics removal. Move memset cleanup
first incase we want to do i8 load store cleanup later.
---
llvm/lib/Target/DirectX/DXILLegalizePass.cpp | 23 +------
llvm/test/CodeGen/DirectX/legalize-memset.ll | 68 ++++++++++++++++----
2 files changed, 58 insertions(+), 33 deletions(-)
diff --git a/llvm/lib/Target/DirectX/DXILLegalizePass.cpp b/llvm/lib/Target/DirectX/DXILLegalizePass.cpp
index d55a84d0126ec..48da829f3f497 100644
--- a/llvm/lib/Target/DirectX/DXILLegalizePass.cpp
+++ b/llvm/lib/Target/DirectX/DXILLegalizePass.cpp
@@ -189,26 +189,6 @@ void emitMemset(IRBuilder<> &Builder, Value *Dst, Value *Val,
}
}
-void removeLifetimesForMemset(CallInst *Memset,
- SmallVectorImpl<Instruction *> &ToRemove) {
- assert(Memset->getCalledFunction()->getIntrinsicID() == Intrinsic::memset &&
- "Expected a memset intrinsic");
-
- Value *DstPtr = Memset->getArgOperand(0);
- DstPtr = DstPtr->stripPointerCasts();
-
- for (User *U : DstPtr->users()) {
- if (auto *CI = dyn_cast<CallInst>(U)) {
- switch (CI->getIntrinsicID()) {
- case Intrinsic::lifetime_start:
- case Intrinsic::lifetime_end:
- ToRemove.push_back(CI);
- break;
- }
- }
- }
-}
-
static void removeMemSet(Instruction &I,
SmallVectorImpl<Instruction *> &ToRemove,
DenseMap<Value *, Value *>) {
@@ -222,7 +202,6 @@ static void removeMemSet(Instruction &I,
dyn_cast<ConstantInt>(CI->getArgOperand(2));
assert(Size && "Expected Size to be a ConstantInt");
emitMemset(Builder, Dst, Val, Size);
- removeLifetimesForMemset(CI, ToRemove);
ToRemove.push_back(CI);
}
}
@@ -255,10 +234,10 @@ class DXILLegalizationPipeline {
LegalizationPipeline;
void initializeLegalizationPipeline() {
+ LegalizationPipeline.push_back(removeMemSet);
LegalizationPipeline.push_back(fixI8TruncUseChain);
LegalizationPipeline.push_back(downcastI64toI32InsertExtractElements);
LegalizationPipeline.push_back(legalizeFreeze);
- LegalizationPipeline.push_back(removeMemSet);
}
};
diff --git a/llvm/test/CodeGen/DirectX/legalize-memset.ll b/llvm/test/CodeGen/DirectX/legalize-memset.ll
index 2d9b28753ef3d..3bef92fe43e09 100644
--- a/llvm/test/CodeGen/DirectX/legalize-memset.ll
+++ b/llvm/test/CodeGen/DirectX/legalize-memset.ll
@@ -1,14 +1,17 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
-; RUN: opt -S -passes='dxil-legalize' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
+; RUN: opt -S -dxil-legalize -dxil-finalize-linkage -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
-define void @replace_float_memset_test() {
-; CHECK-LABEL: define void @replace_float_memset_test() {
+define void @replace_float_memset_test() #0 {
+; CHECK-LABEL: define void @replace_float_memset_test(
+; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [2 x float], align 4
+; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 8, ptr nonnull [[ACCUM_I_FLAT]])
; CHECK-NEXT: [[GEP:%.*]] = getelementptr float, ptr [[ACCUM_I_FLAT]], i32 0
; CHECK-NEXT: store float 0.000000e+00, ptr [[GEP]], align 4
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr float, ptr [[ACCUM_I_FLAT]], i32 1
; CHECK-NEXT: store float 0.000000e+00, ptr [[GEP1]], align 4
+; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 8, ptr nonnull [[ACCUM_I_FLAT]])
; CHECK-NEXT: ret void
;
%accum.i.flat = alloca [2 x float], align 4
@@ -18,13 +21,16 @@ define void @replace_float_memset_test() {
ret void
}
-define void @replace_half_memset_test() {
-; CHECK-LABEL: define void @replace_half_memset_test() {
+define void @replace_half_memset_test() #0 {
+; CHECK-LABEL: define void @replace_half_memset_test(
+; CHECK-SAME: ) #[[ATTR0]] {
; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [2 x half], align 4
+; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr nonnull [[ACCUM_I_FLAT]])
; CHECK-NEXT: [[GEP:%.*]] = getelementptr half, ptr [[ACCUM_I_FLAT]], i32 0
; CHECK-NEXT: store half 0xH0000, ptr [[GEP]], align 2
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr half, ptr [[ACCUM_I_FLAT]], i32 1
; CHECK-NEXT: store half 0xH0000, ptr [[GEP1]], align 2
+; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 4, ptr nonnull [[ACCUM_I_FLAT]])
; CHECK-NEXT: ret void
;
%accum.i.flat = alloca [2 x half], align 4
@@ -34,13 +40,16 @@ define void @replace_half_memset_test() {
ret void
}
-define void @replace_double_memset_test() {
-; CHECK-LABEL: define void @replace_double_memset_test() {
+define void @replace_double_memset_test() #0 {
+; CHECK-LABEL: define void @replace_double_memset_test(
+; CHECK-SAME: ) #[[ATTR0]] {
; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [2 x double], align 4
+; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 16, ptr nonnull [[ACCUM_I_FLAT]])
; CHECK-NEXT: [[GEP:%.*]] = getelementptr double, ptr [[ACCUM_I_FLAT]], i32 0
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP]], align 8
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr double, ptr [[ACCUM_I_FLAT]], i32 1
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP1]], align 8
+; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 16, ptr nonnull [[ACCUM_I_FLAT]])
; CHECK-NEXT: ret void
;
%accum.i.flat = alloca [2 x double], align 4
@@ -50,13 +59,16 @@ define void @replace_double_memset_test() {
ret void
}
-define void @replace_int16_memset_test() {
-; CHECK-LABEL: define void @replace_int16_memset_test() {
+define void @replace_int16_memset_test() #0 {
+; CHECK-LABEL: define void @replace_int16_memset_test(
+; CHECK-SAME: ) #[[ATTR0]] {
; CHECK-NEXT: [[CACHE_I:%.*]] = alloca [2 x i16], align 2
+; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr nonnull [[CACHE_I]])
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i16, ptr [[CACHE_I]], i32 0
; CHECK-NEXT: store i16 0, ptr [[GEP]], align 2
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr i16, ptr [[CACHE_I]], i32 1
; CHECK-NEXT: store i16 0, ptr [[GEP1]], align 2
+; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 4, ptr nonnull [[CACHE_I]])
; CHECK-NEXT: ret void
;
%cache.i = alloca [2 x i16], align 2
@@ -66,11 +78,14 @@ define void @replace_int16_memset_test() {
ret void
}
-define void @replace_int_memset_test() {
-; CHECK-LABEL: define void @replace_int_memset_test() {
+define void @replace_int_memset_test() #0 {
+; CHECK-LABEL: define void @replace_int_memset_test(
+; CHECK-SAME: ) #[[ATTR0]] {
; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [1 x i32], align 4
+; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr nonnull [[ACCUM_I_FLAT]])
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i32, ptr [[ACCUM_I_FLAT]], i32 0
; CHECK-NEXT: store i32 0, ptr [[GEP]], align 4
+; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 4, ptr nonnull [[ACCUM_I_FLAT]])
; CHECK-NEXT: ret void
;
%accum.i.flat = alloca [1 x i32], align 4
@@ -79,3 +94,34 @@ define void @replace_int_memset_test() {
call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %accum.i.flat)
ret void
}
+
+define void @replace_int_memset_to_var_test() #0 {
+; CHECK-LABEL: define void @replace_int_memset_to_var_test(
+; CHECK-SAME: ) #[[ATTR0]] {
+; CHECK-NEXT: [[ACCUM_I_FLAT:%.*]] = alloca [1 x i32], align 4
+; CHECK-NEXT: [[I:%.*]] = alloca i8, align 4
+; CHECK-NEXT: store i8 1, ptr [[I]], align 1
+; CHECK-NEXT: [[I8_LOAD:%.*]] = load i8, ptr [[I]], align 1
+; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr nonnull [[ACCUM_I_FLAT]])
+; CHECK-NEXT: [[TMP1:%.*]] = zext i8 [[I8_LOAD]] to i32
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr i32, ptr [[ACCUM_I_FLAT]], i32 0
+; CHECK-NEXT: store i32 [[TMP1]], ptr [[GEP]], align 4
+; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 4, ptr nonnull [[ACCUM_I_FLAT]])
+; CHECK-NEXT: ret void
+;
+ %accum.i.flat = alloca [1 x i32], align 4
+ %i = alloca i8, align 4
+ store i8 1, ptr %i
+ %i8.load = load i8, ptr %i
+ call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %accum.i.flat)
+ call void @llvm.memset.p0.i32(ptr nonnull align 4 dereferenceable(8) %accum.i.flat, i8 %i8.load, i32 4, i1 false)
+ call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %accum.i.flat)
+ ret void
+}
+
+attributes #0 = {"hlsl.export"}
+
+
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr captures(none))
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr captures(none))
+declare void @llvm.memset.p0.i32(ptr writeonly captures(none), i8, i32, i1 immarg)
More information about the llvm-commits
mailing list