Skip to content

[Inliner] Preserve alignment of byval arguments #137455

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Apr 26, 2025
Merged

[Inliner] Preserve alignment of byval arguments #137455

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 16 additions & 11 deletions llvm/lib/Transforms/Utils/InlineFunction.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1703,19 +1703,19 @@ static void AddAlignmentAssumptions(CallBase &CB, InlineFunctionInfo &IFI) {
}

static void HandleByValArgumentInit(Type *ByValType, Value *Dst, Value *Src,
Module *M, BasicBlock *InsertBlock,
MaybeAlign SrcAlign, Module *M,
BasicBlock *InsertBlock,
InlineFunctionInfo &IFI,
Function *CalledFunc) {
IRBuilder<> Builder(InsertBlock, InsertBlock->begin());

Value *Size =
Builder.getInt64(M->getDataLayout().getTypeStoreSize(ByValType));

// Always generate a memcpy of alignment 1 here because we don't know
// the alignment of the src pointer. Other optimizations can infer
// better alignment.
CallInst *CI = Builder.CreateMemCpy(Dst, /*DstAlign*/ Align(1), Src,
/*SrcAlign*/ Align(1), Size);
Align DstAlign = Dst->getPointerAlignment(M->getDataLayout());

// Generate a memcpy with the correct alignments.
CallInst *CI = Builder.CreateMemCpy(Dst, DstAlign, Src, SrcAlign, Size);

// The verifier requires that all calls of debug-info-bearing functions
// from debug-info-bearing functions have a debug location (for inlining
Expand Down Expand Up @@ -2629,9 +2629,12 @@ llvm::InlineResult llvm::InlineFunction(CallBase &CB, InlineFunctionInfo &IFI,
struct ByValInit {
Value *Dst;
Value *Src;
MaybeAlign SrcAlign;
Type *Ty;
};
// Keep a list of pair (dst, src) to emit byval initializations.
// Keep a list of tuples (dst, src, src_align) to emit byval
// initializations. Src Alignment is only available though the callbase,
// therefore has to be saved.
SmallVector<ByValInit, 4> ByValInits;

// When inlining a function that contains noalias scope metadata,
Expand Down Expand Up @@ -2661,8 +2664,9 @@ llvm::InlineResult llvm::InlineFunction(CallBase &CB, InlineFunctionInfo &IFI,
&CB, CalledFunc, IFI,
CalledFunc->getParamAlign(ArgNo));
if (ActualArg != *AI)
ByValInits.push_back(
{ActualArg, (Value *)*AI, CB.getParamByValType(ArgNo)});
ByValInits.push_back({ActualArg, (Value *)*AI,
CB.getParamAlign(ArgNo),
CB.getParamByValType(ArgNo)});
}

VMap[&*I] = ActualArg;
Expand Down Expand Up @@ -2712,8 +2716,9 @@ llvm::InlineResult llvm::InlineFunction(CallBase &CB, InlineFunctionInfo &IFI,

// Inject byval arguments initialization.
for (ByValInit &Init : ByValInits)
HandleByValArgumentInit(Init.Ty, Init.Dst, Init.Src, Caller->getParent(),
&*FirstNewBlock, IFI, CalledFunc);
HandleByValArgumentInit(Init.Ty, Init.Dst, Init.Src, Init.SrcAlign,
Caller->getParent(), &*FirstNewBlock, IFI,
CalledFunc);

std::optional<OperandBundleUse> ParentDeopt =
CB.getOperandBundle(LLVMContext::OB_deopt);
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/Inline/byval-align.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ define void @byval_caller(ptr nocapture align 64 %a, ptr %b) #0 {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A1:%.*]] = alloca float, align 128
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr [[A1]])
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[A1]], ptr align 1 [[A]], i64 4, i1 false)
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[A1]], ptr align 128 [[A]], i64 4, i1 false)
; CHECK-NEXT: [[LOAD_I:%.*]] = load float, ptr [[A1]], align 4
; CHECK-NEXT: [[B_IDX_I:%.*]] = getelementptr inbounds float, ptr [[B]], i64 8
; CHECK-NEXT: [[ADD_I:%.*]] = fadd float [[LOAD_I]], 2.000000e+00
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/Inline/byval-with-non-alloca-addrspace.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@ define i64 @foo(ptr %arg) {
; CHECK-SAME: ptr [[ARG:%.*]]) {
; CHECK-NEXT: [[ARG1:%.*]] = alloca [[STRUCT:%.*]], align 8
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 16, ptr [[ARG1]])
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[ARG1]], ptr align 1 [[ARG]], i64 16, i1 false)
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[ARG1]], ptr align 8 [[ARG]], i64 16, i1 false)
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr [[STRUCT]], ptr [[ARG1]], i64 0, i32 1
; CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr [[TMP1]], align 4
; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 16, ptr [[ARG1]])
Expand Down
6 changes: 3 additions & 3 deletions llvm/test/Transforms/Inline/byval.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ define i32 @test1() nounwind {
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr [[STRUCT_SS]], ptr [[S]], i32 0, i32 1
; CHECK-NEXT: store i64 2, ptr [[TMP4]], align 4
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 12, ptr [[S1]])
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[S1]], ptr align 1 [[S]], i64 12, i1 false)
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[S1]], ptr [[S]], i64 12, i1 false)
; CHECK-NEXT: [[TMP1_I:%.*]] = load i32, ptr [[S1]], align 4
; CHECK-NEXT: [[TMP2_I:%.*]] = add i32 [[TMP1_I]], 1
; CHECK-NEXT: store i32 [[TMP2_I]], ptr [[S1]], align 4
Expand Down Expand Up @@ -105,7 +105,7 @@ define void @test3() nounwind {
; CHECK-NEXT: [[S1:%.*]] = alloca [[STRUCT_SS:%.*]], align 64
; CHECK-NEXT: [[S:%.*]] = alloca [[STRUCT_SS]], align 1
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 12, ptr [[S1]])
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[S1]], ptr align 1 [[S]], i64 12, i1 false)
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 64 [[S1]], ptr align 64 [[S]], i64 12, i1 false)
; CHECK-NEXT: call void @g3(ptr align 64 [[S1]]) #[[ATTR0]]
; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 12, ptr [[S1]])
; CHECK-NEXT: ret void
Expand Down Expand Up @@ -158,7 +158,7 @@ define i32 @test5() {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[B:%.*]] = alloca [[STRUCT_S0:%.*]], align 8
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr [[B]])
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[B]], ptr align 1 @b, i64 4, i1 false)
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[B]], ptr align 4 @b, i64 4, i1 false)
; CHECK-NEXT: store i32 0, ptr @b, align 4
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[B]], align 4
; CHECK-NEXT: store i32 [[TMP0]], ptr @a, align 4
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/Inline/inline-deferred-instsimplify.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,7 +53,7 @@ define i32 @main() {
; CHECK-LABEL: define i32 @main() {
; CHECK-NEXT: [[G_VAR:%.*]] = alloca [[STRUCT_A:%.*]], align 8
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 20, ptr [[G_VAR]])
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[G_VAR]], ptr align 1 @g_var, i64 20, i1 false)
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[G_VAR]], ptr align 8 @g_var, i64 20, i1 false)
; CHECK-NEXT: [[VAL_I:%.*]] = load i32, ptr [[G_VAR]], align 8
; CHECK-NEXT: [[DOTNOT_I:%.*]] = icmp eq i32 [[VAL_I]], 0
; CHECK-NEXT: br i1 [[DOTNOT_I]], label [[CHECK_POINTERS_ARE_EQUAL_I:%.*]], label [[STORE_PTR_IN_GVAR_I:%.*]]
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/Transforms/Inline/inline-tail.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,7 +65,7 @@ define void @test_byval_a(ptr byval(i32) %p) {
; CHECK-SAME: (ptr byval(i32) [[P:%.*]]) {
; CHECK-NEXT: [[P1:%.*]] = alloca i32, align 4
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr [[P1]])
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[P1]], ptr align 1 [[P]], i64 4, i1 false)
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[P1]], ptr [[P]], i64 4, i1 false)
; CHECK-NEXT: musttail call void @test_byval_c(ptr byval(i32) [[P1]])
; CHECK-NEXT: ret void
;
Expand All @@ -88,7 +88,7 @@ define void @test_dynalloca_a(ptr byval(i32) %p, i32 %n) {
; CHECK-NEXT: [[P1:%.*]] = alloca i32, align 4
; CHECK-NEXT: [[SAVEDSTACK:%.*]] = call ptr @llvm.stacksave.p0()
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 4, ptr [[P1]])
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[P1]], ptr align 1 [[P]], i64 4, i1 false)
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[P1]], ptr [[P]], i64 4, i1 false)
; CHECK-NEXT: [[BUF_I:%.*]] = alloca i8, i32 [[N]], align 1
; CHECK-NEXT: call void @escape(ptr [[BUF_I]])
; CHECK-NEXT: musttail call void @test_dynalloca_c(ptr byval(i32) [[P1]], i32 [[N]])
Expand Down
Loading