Skip to content
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

Jump to bottom

Fix assertion on splat of groupshared scalar #6930

Merged
merged 2 commits into from
Sep 25, 2024
Merged

Fix assertion on splat of groupshared scalar #6930

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
24aecf5
Fix assertion on splat of groupshared scalar
amaiorano Sep 24, 2024
4a80f6b
Add verification to the HLSL test
amaiorano Sep 24, 2024
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
15 changes: 13 additions & 2 deletions tools/clang/lib/Sema/SemaHLSL.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8444,8 +8444,19 @@ ExprResult HLSLExternalSource::LookupVectorMemberExprForHLSL(
ExprValueKind VK = positions.ContainsDuplicateElements()
? VK_RValue
: (IsArrow ? VK_LValue : BaseExpr.getValueKind());
HLSLVectorElementExpr *vectorExpr = new (m_context) HLSLVectorElementExpr(
resultType, VK, &BaseExpr, *member, MemberLoc, positions);

Expr *E = &BaseExpr;
// Insert an lvalue-to-rvalue cast if necessary
if (BaseExpr.getValueKind() == VK_LValue && VK == VK_RValue) {
// Remove qualifiers from result type and cast target type
resultType = resultType.getUnqualifiedType();
auto targetType = E->getType().getUnqualifiedType();
E = ImplicitCastExpr::Create(*m_context, targetType,
CastKind::CK_LValueToRValue, E, nullptr,
VK_RValue);
}
HLSLVectorElementExpr *vectorExpr = new (m_context)
HLSLVectorElementExpr(resultType, VK, E, *member, MemberLoc, positions);

return vectorExpr;
}
Expand Down
15 changes: 15 additions & 0 deletions tools/clang/test/CodeGenHLSL/SplatGroupSharedScalar.hlsl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,15 @@
// RUN: %dxc -E main -T cs_6_0 -fcgl %s | FileCheck %s

// Validate that when swizzling requires an r-value (i.e. duplicate elements), that the result is stored to
// and loaded from a temporary.
// CHECK: store <1 x i32> %splat.splat, <1 x i32>* %tmp
// CHECK-NEXT: %1 = load <1 x i32>, <1 x i32>* %tmp
// CHECK-NEXT: %2 = shufflevector <1 x i32> %1, <1 x i32> undef, <4 x i32> zeroinitializer
// CHECK-NEXT: store <4 x i32> %2, <4 x i32>* %x

groupshared int a;
[numthreads(64, 1, 1)]
void main() {
a = 123;
int4 x = (a).xxxx;
amaiorano marked this conversation as resolved.
Show resolved Hide resolved
}
12 changes: 6 additions & 6 deletions tools/clang/test/CodeGenSPIRV/op.vector.swizzle.hlsl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,18 +118,18 @@ void main() {
// Keep lhs.1
// So final selectors to write to lhs.(0, 1, 2, 3): 6, 1, 4, 5
// CHECK-NEXT: [[v22:%[0-9]+]] = OpLoad %v2float %v2f
// CHECK-NEXT: [[vs15:%[0-9]+]] = OpVectorShuffle %v3float [[v22]] [[v22]] 0 1 0
// CHECK-NEXT: [[vs15:%[0-9]+]] = OpVectorShuffle %v2float [[v22]] [[v22]] 1 0
// CHECK-NEXT: [[vs16:%[0-9]+]] = OpVectorShuffle %v3float [[vs15]] [[vs15]] 1 0 1
// CHECK-NEXT: [[v23:%[0-9]+]] = OpLoad %v4float %v4f2
// CHECK-NEXT: [[vs16:%[0-9]+]] = OpVectorShuffle %v4float [[v23]] [[vs15]] 6 1 4 5
// CHECK-NEXT: OpStore %v4f2 [[vs16]]
// CHECK-NEXT: [[vs17:%[0-9]+]] = OpVectorShuffle %v4float [[v23]] [[vs16]] 6 1 4 5
// CHECK-NEXT: OpStore %v4f2 [[vs17]]
v4f2.wzx.grb = v2f.gr.yxy; // select more than original, write to a part

// CHECK-NEXT: [[v24:%[0-9]+]] = OpLoad %v4float %v4f1
// CHECK-NEXT: OpStore %v4f2 [[v24]]
v4f2.wzyx.abgr.xywz.rgab = v4f1.xyzw.xyzw.rgab.rgab; // from original vector to original vector

// CHECK-NEXT: [[v24_0:%[0-9]+]] = OpAccessChain %_ptr_Function_float %v4f1 %int_2
// CHECK-NEXT: [[ce1:%[0-9]+]] = OpLoad %float [[v24_0]]
// CHECK-NEXT: [[v24_0:%[0-9]+]] = OpLoad %v4float %v4f1
// CHECK-NEXT: [[ce1:%[0-9]+]] = OpCompositeExtract %float [[v24_0]] 2
// CHECK-NEXT: [[ac4:%[0-9]+]] = OpAccessChain %_ptr_Function_float %v4f2 %int_1
// CHECK-NEXT: OpStore [[ac4]] [[ce1]]
v4f2.wzyx.zy.x = v4f1.xzyx.y.x; // from one element (rvalue) to one element (lvalue)
Expand Down
8 changes: 4 additions & 4 deletions tools/clang/test/CodeGenSPIRV/op.vector.swizzle.size1.hlsl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,11 +43,11 @@ void main(float4 input: INPUT) {

// Selecting from resources
// CHECK: [[fptr:%[0-9]+]] = OpAccessChain %_ptr_Uniform_v4float %PerFrame %int_0 %uint_5 %int_0
// CHECK-NEXT: [[elem:%[0-9]+]] = OpAccessChain %_ptr_Uniform_float [[fptr]] %int_3
// CHECK-NEXT: {{%[0-9]+}} = OpLoad %float [[elem]]
// CHECK-NEXT: [[val:%[0-9]+]] = OpLoad %v4float [[fptr]]
// CHECK-NEXT: {{%[0-9]+}} = OpCompositeExtract %float [[val]] 3
v4f = input * PerFrame[5].f.www.r;
// CHECK: [[fptr_0:%[0-9]+]] = OpAccessChain %_ptr_Uniform_v4float %PerFrame %int_0 %uint_6 %int_0
// CHECK-NEXT: [[elem_0:%[0-9]+]] = OpAccessChain %_ptr_Uniform_float [[fptr_0]] %int_2
// CHECK-NEXT: {{%[0-9]+}} = OpLoad %float [[elem_0]]
// CHECK-NEXT: [[val_0:%[0-9]+]] = OpLoad %v4float [[fptr_0]]
// CHECK-NEXT: {{%[0-9]+}} = OpCompositeExtract %float [[val_0]] 2
sf = PerFrame[6].f.zzz.r * input.y;
}
Loading