mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-06-11 16:59:10 +00:00
0bad76b1de
[CK] Remove duplicated XDL/WMMA tests (#4415) ## Motivation When we started the RDNA4 support, the XDL instances were not supporting WMMA instructions, so we duplicated some tests. In this issue, we simplified most of the duplicated test files into common test files. ## Technical Details The following tests were unified: - `batched_gemm` - `batched_gemm_gemm` - `gemm_add` - `gemm_universal` - `grouped_convnd_bwd_data` The following tests were duplicated exactly, and copied into two files with `_xdl` and `_wmma` suffixes. Now they are unified in one single file without suffix: - `gemm_multi_abd` - `gemm_b_scale` There is still an apparent duplication which is a special case, namely `test_grouped_convnd_bwd_weight_interface_{suffix}` where `{suffix}` is `xdl` or `wmma`. However, the WMMA code relies on an old implementation, and is expected to be removed in the future. In addition, it differs from the XDL implementation significantly. Therefore, it was decided to keep both files separate instead of attempting any unification. ## Test Plan `CMakeLists.txt` files were modified to support the new, unified tests. In particular, testing was done for `gfx90a`, `gfx1201` and `gfx11` architectures. ## Test Result All tests passed successfully on all three tested architectures. ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. --------- Co-authored-by: Fernando Jiménez <fernando.jimenez@streamhpc.com>
119 lines
3.8 KiB
C++
119 lines
3.8 KiB
C++
// Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
|
|
// SPDX-License-Identifier: MIT
|
|
|
|
#include <tuple>
|
|
|
|
#include "gtest/gtest.h"
|
|
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
|
|
#include "test_gemm_universal_util.hpp"
|
|
ck::index_t param_mask = 0xffff;
|
|
ck::index_t instance_index = -1;
|
|
using I4 = ck::pk_i4_t;
|
|
using F8 = ck::f8_t;
|
|
using F16 = ck::half_t;
|
|
|
|
using F32 = float;
|
|
|
|
using Row = ck::tensor_layout::gemm::RowMajor;
|
|
using Col = ck::tensor_layout::gemm::ColumnMajor;
|
|
|
|
namespace {
|
|
|
|
template <typename X, typename Y>
|
|
struct tuple_concat;
|
|
|
|
template <typename... Xs, typename... Ys>
|
|
struct tuple_concat<std::tuple<Xs...>, std::tuple<Ys...>>
|
|
{
|
|
using type = std::tuple<Xs..., Ys...>;
|
|
};
|
|
|
|
} // namespace
|
|
|
|
template <typename Tuple>
|
|
class TestGemmUniversal_FP16_MK_KN
|
|
: public ck::test::TestGemmUniversal<typename tuple_concat<std::tuple<Row, Row>, Tuple>::type>
|
|
{
|
|
};
|
|
|
|
template <typename Tuple>
|
|
class TestGemmUniversal_FP16_MK_NK
|
|
: public ck::test::TestGemmUniversal<typename tuple_concat<std::tuple<Row, Col>, Tuple>::type>
|
|
{
|
|
};
|
|
|
|
template <typename Tuple>
|
|
class TestGemmUniversal_FP16_KM_KN
|
|
: public ck::test::TestGemmUniversal<typename tuple_concat<std::tuple<Col, Row>, Tuple>::type>
|
|
{
|
|
};
|
|
|
|
template <typename Tuple>
|
|
class TestGemmUniversal_FP16_KM_NK
|
|
: public ck::test::TestGemmUniversal<typename tuple_concat<std::tuple<Col, Col>, Tuple>::type>
|
|
{
|
|
};
|
|
|
|
// clang-format off
|
|
using KernelTypes_MK_KN = ::testing::Types<
|
|
// ADataType, BDataType, ComputeDataType, CDataType
|
|
#if defined(CK_ENABLE_FP8) && defined(CK_USE_WMMA_FP8)
|
|
std::tuple< F8, F16, F16, F16>,
|
|
std::tuple< F16, F8, F16, F16>,
|
|
#endif
|
|
std::tuple< F16, F16, F16, F16>
|
|
>;
|
|
|
|
using KernelTypes_MK_NK = ::testing::Types<
|
|
// ADataType, BDataType, ComputeDataType, CDataType
|
|
#if defined(CK_ENABLE_FP8) && defined(CK_USE_WMMA_FP8)
|
|
std::tuple< F8, F16, F16, F16>,
|
|
std::tuple< F16, F8, F16, F16>,
|
|
std::tuple< F16, I4, F16, F16>,
|
|
#endif
|
|
std::tuple< F16, F16, F16, F16>
|
|
>;
|
|
|
|
using KernelTypes_KM_NK = ::testing::Types<
|
|
// ADataType, BDataType, ComputeDataType, CDataType
|
|
#if defined(CK_ENABLE_FP8) && defined(CK_USE_WMMA_FP8)
|
|
std::tuple< F8, F16, F16, F16>,
|
|
std::tuple< F16, F8, F16, F16>,
|
|
std::tuple< F16, I4, F16, F16>,
|
|
#endif
|
|
std::tuple< F16, F16, F16, F16>
|
|
>;
|
|
|
|
using KernelTypes_KM_KN = ::testing::Types<
|
|
// ADataType, BDataType, ComputeDataType, CDataType
|
|
#if defined(CK_ENABLE_FP8) && defined(CK_USE_WMMA_FP8)
|
|
std::tuple< F8, F16, F16, F16>,
|
|
std::tuple< F16, F8, F16, F16>,
|
|
#endif
|
|
std::tuple< F16, F16, F16, F16>
|
|
>;
|
|
// clang-format on
|
|
|
|
TYPED_TEST_SUITE(TestGemmUniversal_FP16_MK_KN, KernelTypes_MK_KN);
|
|
TYPED_TEST_SUITE(TestGemmUniversal_FP16_MK_NK, KernelTypes_MK_NK);
|
|
TYPED_TEST_SUITE(TestGemmUniversal_FP16_KM_NK, KernelTypes_KM_NK);
|
|
TYPED_TEST_SUITE(TestGemmUniversal_FP16_KM_KN, KernelTypes_KM_KN);
|
|
|
|
#include "test_gemm_universal_ut_cases_fp16.inc"
|
|
int main(int argc, char** argv)
|
|
{
|
|
testing::InitGoogleTest(&argc, argv);
|
|
if(argc == 1) {}
|
|
else if(argc == 3)
|
|
{
|
|
param_mask = strtol(argv[1], nullptr, 0);
|
|
instance_index = atoi(argv[2]);
|
|
}
|
|
else
|
|
{
|
|
std::cout << "Usage of " << argv[0] << std::endl;
|
|
std::cout << "Arg1,2: param_mask instance_index(-1 means all)" << std::endl;
|
|
}
|
|
return RUN_ALL_TESTS();
|
|
}
|