Skip to content

[AMDGPU] PromoteAlloca: handle out-of-bounds GEP for shufflevector #139700

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
ro-i merged 3 commits into from
May 21, 2025
Merged

[AMDGPU] PromoteAlloca: handle out-of-bounds GEP for shufflevector #139700

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
dce5aea
[AMDGPU] PromoteAlloca: handle out-of-bounds GEP for shufflevector
ro-i May 21, 2025
04afda3
fix typo
ro-i May 21, 2025
74378e1
Update llvm/test/CodeGen/AMDGPU/promote-alloca-shufflevector.ll
ro-i May 21, 2025
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
4 changes: 3 additions & 1 deletion llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -666,7 +666,9 @@ static Value *promoteAllocaUserToVector(
SmallVector<int> Mask;
for (unsigned Idx = 0; Idx < VectorTy->getNumElements(); ++Idx) {
if (Idx >= DestBegin && Idx < DestBegin + NumCopied) {
Mask.push_back(SrcBegin++);
Mask.push_back(SrcBegin < VectorTy->getNumElements()
? SrcBegin++
: PoisonMaskElem);
} else {
Mask.push_back(Idx);
}
Expand Down
44 changes: 44 additions & 0 deletions llvm/test/CodeGen/AMDGPU/promote-alloca-shufflevector.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,44 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -mtriple amdgcn -passes=amdgpu-promote-alloca-to-vector -S < %s | FileCheck %s

; Skip promote-alloca in case of an index which is known to be out of bounds.

define amdgpu_kernel void @out_of_bounds() {
; CHECK-LABEL: define amdgpu_kernel void @out_of_bounds() {
; CHECK-NEXT: [[PTR:%.*]] = freeze <4 x float> poison
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x float> [[PTR]], <4 x float> poison, <4 x i32> <i32 poison, i32 poison, i32 2, i32 3>
; CHECK-NEXT: ret void
;
%ptr = alloca [4 x float], align 4, addrspace(5)
%elem_ptr = getelementptr [4 x float], ptr addrspace(5) %ptr, i32 0, i32 42
call void @llvm.memcpy.p5.p5.i32(ptr addrspace(5) %ptr, ptr addrspace(5) %elem_ptr, i32 8, i1 false)
ret void
}

define amdgpu_kernel void @memcpy_partially_out_of_bounds() {
; CHECK-LABEL: define amdgpu_kernel void @memcpy_partially_out_of_bounds() {
; CHECK-NEXT: [[PTR:%.*]] = freeze <3 x float> poison
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <3 x float> [[PTR]], <3 x float> poison, <3 x i32> <i32 2, i32 poison, i32 2>
; CHECK-NEXT: ret void
;
%ptr = alloca [3 x float], align 4, addrspace(5)
%elem_ptr = getelementptr [3 x float], ptr addrspace(5) %ptr, i32 0, i32 2
call void @llvm.memcpy.p5.p5.i32(ptr addrspace(5) %ptr, ptr addrspace(5) %elem_ptr, i32 8, i1 false)
ret void
}

define amdgpu_kernel void @in_bounds() {
; CHECK-LABEL: define amdgpu_kernel void @in_bounds() {
; CHECK-NEXT: [[PTR:%.*]] = freeze <4 x float> poison
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x float> [[PTR]], <4 x float> poison, <4 x i32> <i32 2, i32 3, i32 2, i32 3>
; CHECK-NEXT: ret void
;
%ptr = alloca [4 x float], align 4, addrspace(5)
%elem_ptr = getelementptr [4 x float], ptr addrspace(5) %ptr, i32 0, i32 2
call void @llvm.memcpy.p5.p5.i32(ptr addrspace(5) %ptr, ptr addrspace(5) %elem_ptr, i32 8, i1 false)
ret void
}

declare void @llvm.memcpy.p5.p5.i32(ptr addrspace(5) writeonly captures(none), ptr addrspace(5) readonly captures(none), i32, i1 immarg) #0

attributes #0 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }