Skip to content

[DXIL] Add scalarization support for WaveReadLaneAt #112570

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
Oct 16, 2024
Merged

[DXIL] Add scalarization support for WaveReadLaneAt #112570

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
3 changes: 3 additions & 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 @@ -18,6 +18,8 @@ using namespace llvm;
bool DirectXTTIImpl::isTargetIntrinsicWithScalarOpAtArg(Intrinsic::ID ID,
unsigned ScalarOpdIdx) {
switch (ID) {
case Intrinsic::dx_wave_readlane:
return ScalarOpdIdx == 1;
default:
return false;
}
Expand All @@ -28,6 +30,7 @@ bool DirectXTTIImpl::isTargetIntrinsicTriviallyScalarizable(
switch (ID) {
case Intrinsic::dx_frac:
case Intrinsic::dx_rsqrt:
case Intrinsic::dx_wave_readlane:
return true;
default:
return false;
Expand Down
35 changes: 35 additions & 0 deletions llvm/test/CodeGen/DirectX/WaveReadLaneAt-vec.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,35 @@
; RUN: opt -S -scalarizer -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-compute %s | FileCheck %s

; Test that for vector values, WaveReadLaneAt scalarizes and maps down to the
; DirectX op

define noundef <2 x half> @wave_read_lane_v2half(<2 x half> noundef %expr, i32 %idx) {
entry:
; CHECK: call half @dx.op.waveReadLaneAt.f16(i32 117, half %expr.i0, i32 %idx)
; CHECK: call half @dx.op.waveReadLaneAt.f16(i32 117, half %expr.i1, i32 %idx)
%ret = call <2 x half> @llvm.dx.wave.readlane.f16(<2 x half> %expr, i32 %idx)
ret <2 x half> %ret
}

define noundef <3 x i32> @wave_read_lane_v3i32(<3 x i32> noundef %expr, i32 %idx) {
entry:
; CHECK: call i32 @dx.op.waveReadLaneAt.i32(i32 117, i32 %expr.i0, i32 %idx)
; CHECK: call i32 @dx.op.waveReadLaneAt.i32(i32 117, i32 %expr.i1, i32 %idx)
; CHECK: call i32 @dx.op.waveReadLaneAt.i32(i32 117, i32 %expr.i2, i32 %idx)
%ret = call <3 x i32> @llvm.dx.wave.readlane(<3 x i32> %expr, i32 %idx)
ret <3 x i32> %ret
}

define noundef <4 x double> @wave_read_lane_v4f64(<4 x double> noundef %expr, i32 %idx) {
entry:
; CHECK: call double @dx.op.waveReadLaneAt.f64(i32 117, double %expr.i0, i32 %idx)
; CHECK: call double @dx.op.waveReadLaneAt.f64(i32 117, double %expr.i1, i32 %idx)
; CHECK: call double @dx.op.waveReadLaneAt.f64(i32 117, double %expr.i2, i32 %idx)
; CHECK: call double @dx.op.waveReadLaneAt.f64(i32 117, double %expr.i3, i32 %idx)
%ret = call <4 x double> @llvm.dx.wave.readlane(<4 x double> %expr, i32 %idx)
ret <4 x double> %ret
}

declare <2 x half> @llvm.dx.wave.readlane.v2f16(<2 x half>, i32)
declare <3 x i32> @llvm.dx.wave.readlane.v3i32(<3 x i32>, i32)
declare <4 x double> @llvm.dx.wave.readlane.v4f64(<4 x double>, i32)
Loading