Skip to content

[HLSL] Use memcpy for HLSL buffer copies that don't have padding #173093

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

Open
bogner wants to merge 1 commit into
base: users/bogner/pr173078
Choose a base branch
from
+8 −13
Open

[HLSL] Use memcpy for HLSL buffer copies that don't have padding #173093

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
11 changes: 6 additions & 5 deletions clang/lib/CodeGen/CGHLSLRuntime.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1532,11 +1532,12 @@ class HLSLBufferCopyEmitter {
bool emitCopy(QualType CType) {
LayoutTy = HLSLBufferLayoutBuilder(CGF.CGM).layOutType(CType);

// TODO: We should be able to fall back to a regular memcpy if the layout
// type doesn't have any padding, but that runs into issues in the backend
// currently.
//
// See https://github.com/llvm/wg-hlsl/issues/351
// If the layout type matches the original type, we can just fall back to a
// regular memcpy.
llvm::Type *OrigTy = CGF.CGM.getTypes().ConvertTypeForMem(CType);
if (LayoutTy == OrigTy)
return false;

emitCopyAtIndices(LayoutTy, llvm::ConstantInt::get(CGF.SizeTy, 0),
llvm::ConstantInt::get(CGF.SizeTy, 0));
return true;
Expand Down
10 changes: 2 additions & 8 deletions clang/test/CodeGenHLSL/ArrayAssignable.hlsl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,17 +153,11 @@ void arr_assign8() {
C = c1;
}

// TODO: We should be able to just memcpy here.
// See https://github.com/llvm/wg-hlsl/issues/351
// Since everything is aligned on 16 byte boundaries, we just get memcpy.
//
// CHECK-LABEL: define hidden void {{.*}}arr_assign9
// CHECK: [[C:%.*]] = alloca [2 x <4 x i32>], align 16
// CHECK-NEXT: [[V0:%.*]] = getelementptr inbounds [2 x <4 x i32>], ptr [[C]], i32 0
// CHECK-NEXT: [[L0:%.*]] = load <4 x i32>, ptr addrspace(2) @c2, align 16
// CHECK-NEXT: store <4 x i32> [[L0]], ptr [[V0]], align 16
// CHECK-NEXT: [[V1:%.*]] = getelementptr inbounds [2 x <4 x i32>], ptr [[C]], i32 0, i32 1
// CHECK-NEXT: [[L1:%.*]] = load <4 x i32>, ptr addrspace(2) getelementptr inbounds ([2 x <4 x i32>], ptr addrspace(2) @c2, i32 0, i32 1), align 16
// CHECK-NEXT: store <4 x i32> [[L1]], ptr [[V1]], align 16
// CHECK-NEXT: call void @llvm.memcpy.p0.p2.i32(ptr align 16 [[C]], ptr addrspace(2) align 16 @c2, i32 32, i1 false)
// CHECK-NEXT: ret void
void arr_assign9() {
int4 C[2];
Expand Down