ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-02 04:37:02 +00:00
Code Issues Packages Projects Releases Wiki Activity

added wmma multiply_multiply instances

Browse Source
This commit is contained in:
Zoltan Lakatos
2025-06-17 15:03:18 +00:00
parent 881bc3fbb0
commit ac60286ed0
13 changed files with 360 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
test/gemm_add/CMakeLists.txt
View File

@@ -1,24 +1,29 @@
add_gtest_executable(test_gemm_add_xdl test_gemm_add_xdl.cpp)
if(result EQUAL 0)
target_link_libraries(test_gemm_add_xdl PRIVATE utility device_gemm_add_instance)
endif()
# add_gtest_executable(test_gemm_add_xdl test_gemm_add_xdl.cpp)
# if(result EQUAL 0)
# target_link_libraries(test_gemm_add_xdl PRIVATE utility device_gemm_add_instance)
# endif()
add_gtest_executable(test_gemm_add_relu_xdl test_gemm_add_relu_xdl.cpp)
if(result EQUAL 0)
target_link_libraries(test_gemm_add_relu_xdl PRIVATE utility device_gemm_add_instance device_gemm_add_relu_instance)
endif()
# add_gtest_executable(test_gemm_add_relu_xdl test_gemm_add_relu_xdl.cpp)
# if(result EQUAL 0)
# target_link_libraries(test_gemm_add_relu_xdl PRIVATE utility device_gemm_add_instance device_gemm_add_relu_instance)
# endif()
add_gtest_executable(test_gemm_add_silu_xdl test_gemm_add_silu_xdl.cpp)
if(result EQUAL 0)
target_link_libraries(test_gemm_add_silu_xdl PRIVATE utility device_gemm_add_instance device_gemm_add_silu_instance)
endif()
# add_gtest_executable(test_gemm_add_silu_xdl test_gemm_add_silu_xdl.cpp)
# if(result EQUAL 0)
# target_link_libraries(test_gemm_add_silu_xdl PRIVATE utility device_gemm_add_instance device_gemm_add_silu_instance)
# endif()
add_gtest_executable(test_gemm_add_fastgelu_xdl test_gemm_add_fastgelu_xdl.cpp)
if(result EQUAL 0)
target_link_libraries(test_gemm_add_fastgelu_xdl PRIVATE utility device_gemm_add_instance device_gemm_add_fastgelu_instance)
endif()
# add_gtest_executable(test_gemm_add_fastgelu_xdl test_gemm_add_fastgelu_xdl.cpp)
# if(result EQUAL 0)
# target_link_libraries(test_gemm_add_fastgelu_xdl PRIVATE utility device_gemm_add_instance device_gemm_add_fastgelu_instance)
# endif()
add_gtest_executable(test_gemm_add_fastgelu_wmma test_gemm_add_fastgelu_wmma.cpp)
# add_gtest_executable(test_gemm_add_fastgelu_wmma test_gemm_add_fastgelu_wmma.cpp)
# if(result EQUAL 0)
# target_link_libraries(test_gemm_add_fastgelu_wmma PRIVATE utility device_gemm_add_fastgelu_instance)
# endif()
add_gtest_executable(test_gemm_multiply_multiply_wmma test_gemm_multiply_multiply_wmma.cpp)
if(result EQUAL 0)
target_link_libraries(test_gemm_add_fastgelu_wmma PRIVATE utility device_gemm_add_fastgelu_instance)
target_link_libraries(test_gemm_multiply_multiply_wmma PRIVATE utility device_gemm_multiply_multiply_instance)
endif()

1
test/gemm_add/test_gemm_common.hpp
View File

@@ -12,6 +12,7 @@ using I8 = int8_t;
using BF16 = ck::bhalf_t;
using F16 = ck::half_t;
using F32 = float;
using I32 = int32_t;
template <typename Tuple>
class TestGemmD0Common : public ::testing::Test

82
test/gemm_add/test_gemm_multiply_multiply_wmma.cpp Normal file
View File

@@ -0,0 +1,82 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
#include "gtest/gtest.h"
#include "ck/ck.hpp"
#include "profiler/profile_gemm_multiply_multiply_impl.hpp"
using Row = ck::tensor_layout::gemm::RowMajor;
using Col = ck::tensor_layout::gemm::ColumnMajor;
using I8 = int8_t;
using BF16 = ck::bhalf_t;
using F16 = ck::half_t;
using F32 = float;
using I32 = int32_t;
template <typename Tuple>
class TestGemmMultiplyMultiply : public ::testing::Test
{
private:
using ADataType = std::tuple_element_t<0, Tuple>;
using BDataType = std::tuple_element_t<1, Tuple>;
using AccDataType = std::tuple_element_t<2, Tuple>;
using D0DataType = std::tuple_element_t<3, Tuple>;
using D1DataType = std::tuple_element_t<4, Tuple>;
using EDataType = std::tuple_element_t<5, Tuple>;
using ALayout = std::tuple_element_t<6, Tuple>;
using BLayout = std::tuple_element_t<7, Tuple>;
using D0Layout = std::tuple_element_t<8, Tuple>;
using D1Layout = std::tuple_element_t<9, Tuple>;
using ELayout = std::tuple_element_t<10, Tuple>;
constexpr static auto ProfileGemmMultiplyMultiplyImpl =
ck::profiler::profile_gemm_multiply_multiply_impl<ADataType,
BDataType,
ADataType, // ComputeDataType
AccDataType,
D0DataType,
D1DataType,
EDataType,
ALayout,
BLayout,
D0Layout,
D1Layout,
ELayout>;
public:
void Run()
{
std::vector<std::vector<ck::index_t>> lengths = {{1024, 1024, 128}};
// std::vector<std::vector<ck::index_t>> lengths = {
// {16, 32, 64}, /*{2048, 4096, 8192},*/ {2048, 4096, 128}};
bool all_success = true;
for(auto length : lengths)
{
int M = length[0];
int N = length[1];
int K = length[2];
int StrideA = ck::is_same_v<ALayout, Row> ? K : M;
int StrideB = ck::is_same_v<BLayout, Row> ? N : K;
int StrideD0 = ck::is_same_v<D0Layout, Row> ? N : M;
int StrideD1 = ck::is_same_v<D1Layout, Row> ? N : M;
int StrideE = ck::is_same_v<ELayout, Row> ? N : M;
all_success =
all_success &
ProfileGemmMultiplyMultiplyImpl(1, 1, false, false, M, N, K, StrideA, StrideB, StrideD0, StrideD1, StrideE, 1, 1, 1, 0);
}
EXPECT_TRUE(all_success);
}
};
using KernelTypes =
::testing::Types<std::tuple<I8, I8, I32, F16, F16, F16, Row, Col, Row, Col, Row>/*,
std::tuple<I8, I8, I32, F32, F32, BF16, Row, Col, Row, Col, Row>*/>;
TYPED_TEST_SUITE(TestGemmMultiplyMultiply, KernelTypes);
TYPED_TEST(TestGemmMultiplyMultiply, Test_BF16FP16) { this->Run(); }