Doomhowl-Interactive/llvm-project
0
0
Fork 0
Code Issues Pull Requests Actions 5 Packages Projects Releases Wiki Activity

[HLSL] Allow 1x1 matrices to be splatted like scalars (#188119)

Browse Source 
Fixes #186859 by allowing 1x1 matrices to be splatted like the scalar
and vec1 cases.

Assisted-by: GitHub Copilot (powered by Claude Opus 4.6)
This commit is contained in:
Deric C.
2026-03-23 14:49:53 -07:00
committed by GitHub
parent 599d74fdc3
commit bcfe1e94d3
3 changed files with 48 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
clang/lib/Sema/SemaCast.cpp
View File

@@ -2939,11 +2939,17 @@ bool CastOperation::CheckHLSLCStyleCast(CheckedConversionKind CCK) {
if (Self.HLSL().CanPerformAggregateSplatCast(SrcExpr.get(), DestType)) {
SrcExpr = Self.DefaultLvalueConversion(SrcExpr.get());
const VectorType *VT = SrcTy->getAs<VectorType>();
const ConstantMatrixType *MT = SrcTy->getAs<ConstantMatrixType>();
// change splat from vec1 case to splat from scalar
if (VT && VT->getNumElements() == 1)
SrcExpr = Self.ImpCastExprToType(
SrcExpr.get(), VT->getElementType(), CK_HLSLVectorTruncation,
SrcExpr.get()->getValueKind(), nullptr, CCK);
// change splat from 1x1 matrix case to splat from scalar
else if (MT && MT->getNumElementsFlattened() == 1)
SrcExpr = Self.ImpCastExprToType(
SrcExpr.get(), MT->getElementType(), CK_HLSLMatrixTruncation,
SrcExpr.get()->getValueKind(), nullptr, CCK);
// Inserting a scalar cast here allows for a simplified codegen in
// the case the destTy is a vector
if (const VectorType *DVT = DestType->getAs<VectorType>())

13
clang/lib/Sema/SemaHLSL.cpp
View File

@@ -4644,8 +4644,8 @@ bool SemaHLSL::CanPerformScalarCast(QualType SrcTy, QualType DestTy) {
}
// Can perform an HLSL Aggregate splat cast if the Dest is an aggregate and the
// Src is a scalar or a vector of length 1
// Or if Dest is a vector and Src is a vector of length 1
// Src is a scalar, a vector of length 1, or a 1x1 matrix
// Or if Dest is a vector and Src is a vector of length 1 or a 1x1 matrix
bool SemaHLSL::CanPerformAggregateSplatCast(Expr *Src, QualType DestTy) {
QualType SrcTy = Src->getType();
@@ -4656,13 +4656,18 @@ bool SemaHLSL::CanPerformAggregateSplatCast(Expr *Src, QualType DestTy) {
return false;
const VectorType *SrcVecTy = SrcTy->getAs<VectorType>();
const ConstantMatrixType *SrcMatTy = SrcTy->getAs<ConstantMatrixType>();
// Src isn't a scalar or a vector of length 1
if (!SrcTy->isScalarType() && !(SrcVecTy && SrcVecTy->getNumElements() == 1))
// Src isn't a scalar, a vector of length 1, or a 1x1 matrix
if (!SrcTy->isScalarType() &&
!(SrcVecTy && SrcVecTy->getNumElements() == 1) &&
!(SrcMatTy && SrcMatTy->getNumElementsFlattened() == 1))
return false;
if (SrcVecTy)
SrcTy = SrcVecTy->getElementType();
else if (SrcMatTy)
SrcTy = SrcMatTy->getElementType();
llvm::SmallVector<QualType> DestTypes;
BuildFlattenedTypeList(DestTy, DestTypes);

33
clang/test/CodeGenHLSL/BasicFeatures/AggregateSplatCast.hlsl
View File

@@ -84,6 +84,39 @@ export void call5() {
S s = (S)A;
}
// vector splat from 1x1 matrix
// CHECK-LABEL: define void {{.*}}call9
// CHECK: [[M:%.*]] = alloca [1 x <1 x float>], align 4
// CHECK-NEXT: [[A:%.*]] = alloca <4 x i32>, align 4
// CHECK-NEXT: store <1 x float> {{.*}}, ptr [[M]], align 4
// CHECK-NEXT: [[L:%.*]] = load <1 x float>, ptr [[M]], align 4
// CHECK-NEXT: [[ML:%.*]] = extractelement <1 x float> [[L]], i32 0
// CHECK-NEXT: [[C:%.*]] = fptosi float [[ML]] to i32
// CHECK-NEXT: [[SI:%.*]] = insertelement <4 x i32> poison, i32 [[C]], i64 0
// CHECK-NEXT: [[S:%.*]] = shufflevector <4 x i32> [[SI]], <4 x i32> poison, <4 x i32> zeroinitializer
// CHECK-NEXT: store <4 x i32> [[S]], ptr [[A]], align 4
export void call9() {
float1x1 M = {1.0};
int4 A = (int4)M;
}
// struct splat from 1x1 matrix
// CHECK-LABEL: define void {{.*}}call10
// CHECK: [[M:%.*]] = alloca [1 x <1 x i32>], align 4
// CHECK-NEXT: [[s:%.*]] = alloca %struct.S, align 1
// CHECK-NEXT: store <1 x i32> splat (i32 1), ptr [[M]], align 4
// CHECK-NEXT: [[L:%.*]] = load <1 x i32>, ptr [[M]], align 4
// CHECK-NEXT: [[ML:%.*]] = extractelement <1 x i32> [[L]], i32 0
// CHECK-NEXT: [[G1:%.*]] = getelementptr inbounds %struct.S, ptr [[s]], i32 0, i32 0
// CHECK-NEXT: [[G2:%.*]] = getelementptr inbounds %struct.S, ptr [[s]], i32 0, i32 1
// CHECK-NEXT: store i32 [[ML]], ptr [[G1]], align 4
// CHECK-NEXT: [[C:%.*]] = sitofp i32 [[ML]] to float
// CHECK-NEXT: store float [[C]], ptr [[G2]], align 4
export void call10() {
int1x1 M = {1};
S s = (S)M;
}
struct BFields {
double DF;
int E: 15;