Skip to content

[DirectX] Scalarize the dx.saturate intrinsic #134381

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 7, 2025
Merged

[DirectX] Scalarize the dx.saturate intrinsic #134381

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
1 change: 1 addition & 0 deletions llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,7 @@ bool DirectXTTIImpl::isTargetIntrinsicTriviallyScalarizable(
case Intrinsic::dx_firstbituhigh:
case Intrinsic::dx_frac:
case Intrinsic::dx_rsqrt:
case Intrinsic::dx_saturate:
case Intrinsic::dx_splitdouble:
case Intrinsic::dx_wave_readlane:
case Intrinsic::dx_wave_reduce_max:
Expand Down
36 changes: 31 additions & 5 deletions llvm/test/CodeGen/DirectX/saturate.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-library %s | FileCheck %s
; RUN: opt -S -scalarizer -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
; Make sure the intrinsic dx.saturate is to appropriate DXIL op for half/float/double data types.

; CHECK-LABEL: test_saturate_half
Expand Down Expand Up @@ -28,9 +28,35 @@ entry:
ret double %hlsl.saturate
}

; CHECK: attributes #[[#ATTR]] = {{{.*}} memory(none) {{.*}}}
; CHECK-LABEL: test_saturate_half4
define noundef <4 x half> @test_saturate_half4(<4 x half> noundef %p0) {
entry:
; CHECK: call half @dx.op.unary.f16(i32 7, half
; CHECK: call half @dx.op.unary.f16(i32 7, half
; CHECK: call half @dx.op.unary.f16(i32 7, half
; CHECK: call half @dx.op.unary.f16(i32 7, half
%hlsl.saturate = call <4 x half> @llvm.dx.saturate.v4f16(<4 x half> %p0)
ret <4 x half> %hlsl.saturate
}

; CHECK-LABEL: test_saturate_float3
define noundef <3 x float> @test_saturate_float3(<3 x float> noundef %p0) {
entry:
; CHECK: call float @dx.op.unary.f32(i32 7, float
; CHECK: call float @dx.op.unary.f32(i32 7, float
; CHECK: call float @dx.op.unary.f32(i32 7, float
%hlsl.saturate = call <3 x float> @llvm.dx.saturate.v3f32(<3 x float> %p0)
ret <3 x float> %hlsl.saturate
}

declare half @llvm.dx.saturate.f16(half)
declare float @llvm.dx.saturate.f32(float)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure why we removed these. Were they causing an error? If you want to make sure they are gone we could do a CHECK-NOT: for these intrinsics?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

They're just unnecessary noise. The IR parser will add declarations of intrinsics when it sees uses, so rather than list every overload we happen to use in the file we can just let the IR parser create them lazily.

CHECK-NOT wouldn't be appropriate here, since they will correctly be present in the output.

declare double @llvm.dx.saturate.f64(double)
; CHECK-LABEL: test_saturate_double2
define noundef <2 x double> @test_saturate_double2(<2 x double> noundef %p0) {
entry:
; CHECK: call double @dx.op.unary.f64(i32 7, double
; CHECK: call double @dx.op.unary.f64(i32 7, double
%hlsl.saturate = call <2 x double> @llvm.dx.saturate.v4f64(<2 x double> %p0)
ret <2 x double> %hlsl.saturate
}


; CHECK: attributes #[[#ATTR]] = {{{.*}} memory(none) {{.*}}}
Loading