Skip to content

[SM6.10][Bugfix] Fix Size check for input interpreted vector in MultiplyAdd #8388

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
hekota merged 2 commits into from
Apr 20, 2026
+67 −2
Merged

[SM6.10][Bugfix] Fix Size check for input interpreted vector in MultiplyAdd #8388

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
4e8bfc9
[SM6.10][Bugfix] Fix Size check for input interpreted vector in Multi…
hekota Apr 18, 2026
4331e28
Update the check to compare VecK against the scalar count calculated …
hekota Apr 20, 2026
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
12 changes: 10 additions & 2 deletions tools/clang/lib/Headers/hlsl/dx/linalg.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -204,6 +204,14 @@ template <SIZE_TYPE MVal, SIZE_TYPE NVal> struct DimMN<MVal, NVal, true> {
static const SIZE_TYPE N = MVal;
};

template <ComponentEnum CompTy, SIZE_TYPE PackedComponentCount>
struct ScalarCountFromPackedComponents {
static const SIZE_TYPE ElementsPerScalar =
ComponentTypeTraits<CompTy>::ElementsPerScalar;
static const SIZE_TYPE Value =
(PackedComponentCount + ElementsPerScalar - 1) / ElementsPerScalar;
};

} // namespace __detail

template <ComponentEnum ElementType, uint DimA> struct VectorRef {
Expand Down Expand Up @@ -506,7 +514,7 @@ template <typename OutputElTy, typename InputElTy, ComponentEnum InputInterp,
ComponentEnum MatrixDT>
// clang-format off
typename hlsl::enable_if<
InterpretedVector<InputElTy, VecK, InputInterp>::Size == K,
VecK == __detail::ScalarCountFromPackedComponents<InputInterp, K>::Value,
vector<OutputElTy, M> >::type
// clang-format on
MultiplyAdd(Matrix<MatrixDT, M, K, MatrixUse::A, MatrixScope::Thread> MatrixA,
Expand Down Expand Up @@ -542,7 +550,7 @@ template <typename OutputElTy, typename InputElTy, ComponentEnum InputInterp,
ComponentEnum MatrixDT>
// clang-format off
typename hlsl::enable_if<
InterpretedVector<InputElTy, VecK, InputInterp>::Size == K,
VecK == __detail::ScalarCountFromPackedComponents<InputInterp, K>::Value,
vector<OutputElTy, M> >::type
// clang-format on
MultiplyAdd(Matrix<MatrixDT, M, K, MatrixUse::A, MatrixScope::Thread> MatrixA,
Expand Down
57 changes: 57 additions & 0 deletions tools/clang/test/CodeGenDXIL/hlsl/linalg/api/vectors.hlsl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,7 @@ using namespace dx::linalg;
using MatrixATy = Matrix<ComponentType::F16, 8, 4, MatrixUse::A, MatrixScope::Thread>;
using MatrixAccum_8_8_Ty = Matrix<ComponentType::F16, 8, 8, MatrixUse::Accumulator, MatrixScope::Thread>;
using MatrixAccum_8_4_Ty = Matrix<ComponentType::F16, 8, 4, MatrixUse::Accumulator, MatrixScope::Thread>;
using Matrix_7_15_ATy = Matrix<ComponentType::F16, 7, 15, MatrixUse::A, MatrixScope::Thread>;

ByteAddressBuffer BAB : register(t0);

Expand Down Expand Up @@ -87,4 +88,60 @@ void main(uint ID : SV_GroupID) {
half3 ThreeF16 = BAB.Load<half3>(256);
InterpretedVector<uint, 1, ComponentEnum::F8_E4M3FN> convertedPacked2 =
Convert<ComponentEnum::F8_E4M3FN, ComponentEnum::F16>(ThreeF16);

// Test MultiplyAdd with odd sizes
//
vector<half, 15> vecH15 = BAB.Load< vector<half, 15> >(168);
vector<half, 7> vecH7 = BAB.Load< vector<half, 7> >(64);

InterpretedVector<half, 15, ComponentEnum::F16> interpVecH15 = MakeInterpretedVector<ComponentEnum::F16>(vecH15);

// CHECK: %[[MAT_7_15:.*]] = call %dx.types.LinAlgMatrixC8M7N15U0S0 @dx.op.linAlgMatrixLoadFromDescriptor.mC8M7N15U0S0(i32 -2147483634,
// CHECK-SAME: %dx.types.Handle %{{[0-9]+}}, i32 0, i32 16, i32 1, i32 128) ; LinAlgMatrixLoadFromDescriptor(handle,offset,stride,layout,align)
Matrix_7_15_ATy Mat_7_15 = Matrix_7_15_ATy::Load<MatrixLayoutEnum::ColMajor>(BAB, 0, 16);

// CHECK: call <7 x half> @dx.op.linAlgMatVecMulAdd.v7f16.mC8M7N15U0S0.v15f16.v7f16(i32 -2147483622,
// CHECK-SAME: %dx.types.LinAlgMatrixC8M7N15U0S0 %[[MAT_7_15]], i1 true, <15 x half> %{{[0-9]+}}, i32 8, <7 x half> %{{[0-9]+}}, i32 8)
// CHECK-SAME: ; LinAlgMatVecMulAdd(matrix,isOutputSigned,inputVector,inputInterpretation,biasVector,biasInterpretation)
vector<half, 7> vec7 = MultiplyAdd<half>(Mat_7_15, vecH15, vecH7);

// CHECK: call <7 x half> @dx.op.linAlgMatVecMulAdd.v7f16.mC8M7N15U0S0.v15f16.v7f16(i32 -2147483622, %dx.types.LinAlgMatrixC8M7N15U0S0 %[[MAT_7_15]],
// CHECK-SAME; i1 true, <15 x half> %{{[0-9]+}}, i32 8, <7 x half> %{{[0-9]+}}, i32 8)
// CHECK-SAME: ; LinAlgMatVecMulAdd(matrix,isOutputSigned,inputVector,inputInterpretation,biasVector,biasInterpretation)
vector<half, 7> vec8 = MultiplyAdd<half>(Mat_7_15, interpVecH15, vecH7);

// CHECK: %[[LOAD1:.*]] = call %dx.types.ResRet.v7f16 @dx.op.rawBufferVectorLoad.v7f16(i32 303, %dx.types.Handle %{{[0-9]+}}, i32 512, i32 undef, i32 2)
// CHECK-SAME: ; RawBufferVectorLoad(buf,index,elementOffset,alignment)
// CHECK: %[[MEM_BIAS1:.*]] = extractvalue %dx.types.ResRet.v7f16 %[[LOAD1]], 0
// CHECK: call <7 x half> @dx.op.linAlgMatVecMulAdd.v7f16.mC8M7N15U0S0.v15f16.v7f16(i32 -2147483622,
// CHECK-SAME: %dx.types.LinAlgMatrixC8M7N15U0S0 %[[MAT_7_15]], i1 true, <15 x half> %29, i32 8, <7 x half> %37, i32 8)
// CHECK-SAME: ; LinAlgMatVecMulAdd(matrix,isOutputSigned,inputVector,inputInterpretation,biasVector,biasInterpretation)
VectorRef<ComponentType::F16, 7> memBias7 = {BAB, 512};
vector<half, 7> vec9 = MultiplyAdd<half>(Mat_7_15, vecH15, memBias7);

// CHECK: %[[LOAD2:.*]] = call %dx.types.ResRet.v7f16 @dx.op.rawBufferVectorLoad.v7f16(i32 303, %dx.types.Handle %{{[0-9]+}}, i32 512, i32 undef, i32 2)
// CHECK-SAME: ; RawBufferVectorLoad(buf,index,elementOffset,alignment)
// CHECK: %[[MEM_BIAS2:.*]] = extractvalue %dx.types.ResRet.v7f16 %[[LOAD2]], 0
// CHECK-NEXT: %dx.types.LinAlgMatrixC8M7N15U0S0 %[[MAT_7_15]], i1 true, <15 x half> %{{[0-9]+}}, i32 8, <7 x half> %[[MEM_BIAS2]], i32 8)
// CHECK-SAME: ; LinAlgMatVecMulAdd(matrix,isOutputSigned,inputVector,inputInterpretation,biasVector,biasInterpretation)
vector<half, 7> vec10 = MultiplyAdd<half>(Mat_7_15, interpVecH15, memBias7);

// Test MultiplyAdd with packed input vector
//
// CHECK: %[[INTERP_VEC_H15_PACKED:.*]] = call <4 x i32> @dx.op.linAlgConvert.v4i32.v15f16(i32 -2147483618,
// CHECK-SAME: <15 x half> %{{[0-9]+}}, i32 8, i32 21) ; LinAlgConvert(inputVector,inputInterpretation,outputInterpretation)
InterpretedVector<uint, 4, ComponentEnum::F8_E4M3FN> interpVecH15Packed = Convert<ComponentEnum::F8_E4M3FN, ComponentEnum::F16>(vecH15);

// CHECK: call <7 x half> @dx.op.linAlgMatVecMulAdd.v7f16.mC8M7N15U0S0.v4i32.v7f16(i32 -2147483622,
// CHECK-SAME: %dx.types.LinAlgMatrixC8M7N15U0S0 %[[MAT_7_15]], i1 true, <4 x i32> %43, i32 21, <7 x half> %31, i32 8)
// CHECK-SAME: ; LinAlgMatVecMulAdd(matrix,isOutputSigned,inputVector,inputInterpretation,biasVector,biasInterpretation)
vector<half, 7> vec11 = MultiplyAdd<half>(Mat_7_15, interpVecH15Packed, vecH7);

// CHECK: %[[LOAD3:.+]] = call %dx.types.ResRet.v7f16 @dx.op.rawBufferVectorLoad.v7f16(i32 303, %dx.types.Handle %45, i32 512, i32 undef, i32 2)
// CHECK-SAME: ; RawBufferVectorLoad(buf,index,elementOffset,alignment)
// CHECK-NEXT: %[[MEM_BIAS3:.*]] = extractvalue %dx.types.ResRet.v7f16 %46, 0
// CHECK-NEXT: call <7 x half> @dx.op.linAlgMatVecMulAdd.v7f16.mC8M7N15U0S0.v4i32.v7f16(i32 -2147483622,
// CHECK-SAME: %dx.types.LinAlgMatrixC8M7N15U0S0 %[[MAT_7_15]], i1 true, <4 x i32> %[[INTERP_VEC_H15_PACKED]], i32 21, <7 x half> %[[MEM_BIAS3]], i32 8)
// CHECK-SAME: ; LinAlgMatVecMulAdd(matrix,isOutputSigned,inputVector,inputInterpretation,biasVector,biasInterpretation)
vector<half, 7> vec12 = MultiplyAdd<half>(Mat_7_15, interpVecH15Packed, memBias7);
}
Loading