Skip to content

[HLSL][DirectX] Fix dot2add DXIL operation to use float overload #152781

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 2 commits into from
Aug 11, 2025
Merged

[HLSL][DirectX] Fix dot2add DXIL operation to use float overload #152781

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4e8c638
fix dot2add definition in DXIL.td and tests
kmpeng Aug 8, 2025
88b605f
add SM 6.3 error test
kmpeng Aug 11, 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
2 changes: 1 addition & 1 deletion clang/test/CodeGenHLSL/builtins/dot2add.hlsl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
// RUN: %clang_cc1 -finclude-default-header -fnative-half-type -triple \
// RUN: dxil-pc-shadermodel6.3-compute %s -emit-llvm -o - | \
// RUN: dxil-pc-shadermodel6.4-compute %s -emit-llvm -o - | \
// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-DXIL
// RUN: %clang_cc1 -finclude-default-header -fnative-half-type -triple \
// RUN: spirv-pc-vulkan-compute %s -emit-llvm -o - | \
Expand Down
10 changes: 5 additions & 5 deletions llvm/lib/Target/DirectX/DXIL.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1108,11 +1108,11 @@ def RawBufferStore : DXILOp<140, rawBufferStore> {
def Dot2AddHalf : DXILOp<162, dot2AddHalf> {
let Doc = "2D half dot product with accumulate to float";
let intrinsics = [IntrinSelect<int_dx_dot2add>];
let arguments = [FloatTy, HalfTy, HalfTy, HalfTy, HalfTy];
let result = FloatTy;
let overloads = [Overloads<DXIL1_0, []>];
let stages = [Stages<DXIL1_0, [all_stages]>];
let attributes = [Attributes<DXIL1_0, [ReadNone]>];
let arguments = [OverloadTy, HalfTy, HalfTy, HalfTy, HalfTy];
let result = OverloadTy;
let overloads = [Overloads<DXIL1_4, [FloatTy]>];
let stages = [Stages<DXIL1_4, [all_stages]>];
let attributes = [Attributes<DXIL1_4, [ReadNone]>];
}

def Dot4AddI8Packed : DXILOp<163, dot4AddPacked> {
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/CodeGen/DirectX/dot2add.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
; RUN: opt -S -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-compute %s | FileCheck %s
; RUN: opt -S -dxil-op-lower -mtriple=dxil-pc-shadermodel6.4-compute %s | FileCheck %s

define noundef float @dot2add_simple(<2 x half> noundef %a, <2 x half> noundef %b, float %acc) {
entry:
Expand All @@ -7,7 +7,7 @@ entry:
%bx = extractelement <2 x half> %b, i32 0
%by = extractelement <2 x half> %b, i32 1

; CHECK: call float @dx.op.dot2AddHalf(i32 162, float %acc, half %ax, half %ay, half %bx, half %by)
; CHECK: call float @dx.op.dot2AddHalf.f32(i32 162, float %acc, half %ax, half %ay, half %bx, half %by)
%ret = call float @llvm.dx.dot2add(float %acc, half %ax, half %ay, half %bx, half %by)
ret float %ret
}
15 changes: 15 additions & 0 deletions llvm/test/CodeGen/DirectX/dot2add_error.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,15 @@
; RUN: not opt -S -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-compute %s 2>&1 | FileCheck %s

; CHECK: in function f
; CHECK-SAME: Cannot create Dot2AddHalf operation: No valid overloads for DXIL version 1.3

define noundef float @f(<2 x half> noundef %a, <2 x half> noundef %b, float %acc) {
entry:
%ax = extractelement <2 x half> %a, i32 0
%ay = extractelement <2 x half> %a, i32 1
%bx = extractelement <2 x half> %b, i32 0
%by = extractelement <2 x half> %b, i32 1

%ret = call float @llvm.dx.dot2add(float %acc, half %ax, half %ay, half %bx, half %by)
ret float %ret
}