Skip to content

[SYCL][Matrix] Add support for joint_matrix_mad when A is uint8/int8 … #4486

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.

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 2 commits into from
Sep 8, 2021
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
36 changes: 36 additions & 0 deletions sycl/include/CL/__spirv/spirv_ops.hpp
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,42 @@ __spirv_JointMatrixMadINTEL(
__spv::__spirv_JointMatrixINTEL<T2, M, N, LC, S> *C,
__spv::Scope::Flag Sc = __spv::Scope::Flag::Subgroup);

template <typename T1, typename T2, typename T3, std::size_t M, std::size_t K,
std::size_t N, __spv::MatrixLayout LA = __spv::MatrixLayout::RowMajor,
__spv::MatrixLayout LB = __spv::MatrixLayout::RowMajor,
__spv::MatrixLayout LC = __spv::MatrixLayout::RowMajor,
__spv::Scope::Flag S = __spv::Scope::Flag::Subgroup>
extern SYCL_EXTERNAL __spv::__spirv_JointMatrixINTEL<T3, M, N, LC, S> *
__spirv_JointMatrixUUMadINTEL(
__spv::__spirv_JointMatrixINTEL<T1, M, K, LA, S> *A,
__spv::__spirv_JointMatrixINTEL<T2, K, N, LB, S> *B,
__spv::__spirv_JointMatrixINTEL<T3, M, N, LC, S> *C,
__spv::Scope::Flag Sc = __spv::Scope::Flag::Subgroup);

template <typename T1, typename T2, typename T3, std::size_t M, std::size_t K,
std::size_t N, __spv::MatrixLayout LA = __spv::MatrixLayout::RowMajor,
__spv::MatrixLayout LB = __spv::MatrixLayout::RowMajor,
__spv::MatrixLayout LC = __spv::MatrixLayout::RowMajor,
__spv::Scope::Flag S = __spv::Scope::Flag::Subgroup>
extern SYCL_EXTERNAL __spv::__spirv_JointMatrixINTEL<T3, M, N, LC, S> *
__spirv_JointMatrixUSMadINTEL(
__spv::__spirv_JointMatrixINTEL<T1, M, K, LA, S> *A,
__spv::__spirv_JointMatrixINTEL<T2, K, N, LB, S> *B,
__spv::__spirv_JointMatrixINTEL<T3, M, N, LC, S> *C,
__spv::Scope::Flag Sc = __spv::Scope::Flag::Subgroup);

template <typename T1, typename T2, typename T3, std::size_t M, std::size_t K,
std::size_t N, __spv::MatrixLayout LA = __spv::MatrixLayout::RowMajor,
__spv::MatrixLayout LB = __spv::MatrixLayout::RowMajor,
__spv::MatrixLayout LC = __spv::MatrixLayout::RowMajor,
__spv::Scope::Flag S = __spv::Scope::Flag::Subgroup>
extern SYCL_EXTERNAL __spv::__spirv_JointMatrixINTEL<T3, M, N, LC, S> *
__spirv_JointMatrixSUMadINTEL(
__spv::__spirv_JointMatrixINTEL<T1, M, K, LA, S> *A,
__spv::__spirv_JointMatrixINTEL<T2, K, N, LB, S> *B,
__spv::__spirv_JointMatrixINTEL<T3, M, N, LC, S> *C,
__spv::Scope::Flag Sc = __spv::Scope::Flag::Subgroup);

#ifndef __SPIRV_BUILTIN_DECLARATIONS__
#error \
"SPIR-V built-ins are not available. Please set -fdeclare-spirv-builtins flag."
Expand Down
25 changes: 18 additions & 7 deletions sycl/include/sycl/ext/oneapi/matrix/matrix-jit.hpp
Original file line number Diff line number Diff line change
Expand Up @@ -160,16 +160,27 @@ joint_matrix_store(Group sg,
#endif // __SYCL_DEVICE_ONLY__
}

template <typename Group, typename T1, typename T2, size_t M, size_t K,
size_t N, matrix_layout LayoutA, matrix_layout LayoutB,
template <typename Group, typename T1, typename T2, typename T3, size_t M,
size_t K, size_t N, matrix_layout LayoutA, matrix_layout LayoutB,
matrix_layout LayoutC>
inline __SYCL_ALWAYS_INLINE joint_matrix<T2, M, N, LayoutC, Group>
inline __SYCL_ALWAYS_INLINE joint_matrix<T3, M, N, LayoutC, Group>
joint_matrix_mad(Group sg, joint_matrix<T1, M, K, LayoutA, Group> &mA,
joint_matrix<T1, K, N, LayoutB, Group> &mB,
joint_matrix<T2, M, N, LayoutC, Group> &mC) {
joint_matrix<T2, K, N, LayoutB, Group> &mB,
joint_matrix<T3, M, N, LayoutC, Group> &mC) {
#ifdef __SYCL_DEVICE_ONLY__
joint_matrix<T2, M, N, LayoutC, Group> res(sg);
res.spvm = __spirv_JointMatrixMadINTEL(mA.spvm, mB.spvm, mC.spvm);
joint_matrix<T3, M, N, LayoutC, Group> res(sg);
if constexpr (std::is_same<T1, uint16_t>::value &&
std::is_same<T2, uint16_t>::value &&
std::is_same<T3, float>::value)
res.spvm = __spirv_JointMatrixMadINTEL(mA.spvm, mB.spvm, mC.spvm);
else if constexpr (std::is_unsigned<T1>::value && std::is_unsigned<T2>::value)
res.spvm = __spirv_JointMatrixUUMadINTEL(mA.spvm, mB.spvm, mC.spvm);
else if constexpr (std::is_signed<T1>::value && std::is_unsigned<T2>::value)
res.spvm = __spirv_JointMatrixSUMadINTEL(mA.spvm, mB.spvm, mC.spvm);
else if constexpr (std::is_unsigned<T1>::value && std::is_signed<T2>::value)
res.spvm = __spirv_JointMatrixUSMadINTEL(mA.spvm, mB.spvm, mC.spvm);
else
res.spvm = __spirv_JointMatrixMadINTEL(mA.spvm, mB.spvm, mC.spvm);
return res;
#else
(void)sg;
Expand Down
2 changes: 1 addition & 1 deletion sycl/test/matrix/matrix-bf16-test-SG-16.cpp
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
// RUN: %clangxx -march=sapphirerapids -fsycl -O2 %s -o %t.out
// RUN: %clangxx -fsycl -O2 %s -o %t.out
#include <CL/sycl.hpp>
#if (SYCL_EXT_ONEAPI_MATRIX == 2)
#include <iostream>
Expand Down
2 changes: 1 addition & 1 deletion sycl/test/matrix/matrix-bf16-test.cpp
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
// RUN: %clangxx -march=sapphirerapids -fsycl -O2 %s -o %t.out
// RUN: %clangxx -fsycl -O2 %s -o %t.out
#include <CL/sycl.hpp>
#if (SYCL_EXT_ONEAPI_MATRIX == 2)
#include <iostream>
Expand Down
2 changes: 1 addition & 1 deletion sycl/test/matrix/matrix-int8-test-SG-16.cpp
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
// RUN: %clangxx -march=sapphirerapids -fsycl -O2 %s -o %t.out
// RUN: %clangxx -fsycl -O2 %s -o %t.out
#include <CL/sycl.hpp>
#if (SYCL_EXT_ONEAPI_MATRIX == 2)
#include <iostream>
Expand Down
2 changes: 1 addition & 1 deletion sycl/test/matrix/matrix-int8-test.cpp
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
// RUN: %clangxx -march=sapphirerapids -fsycl -O2 %s -o %t.out
// RUN: %clangxx -fsycl -O2 %s -o %t.out
#include <CL/sycl.hpp>
#if (SYCL_EXT_ONEAPI_MATRIX == 2)
#include <iostream>
Expand Down