mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-06-29 11:16:59 +00:00
717f2efef7
[CK] add composable kernel support on gfx1250 (#6978) ## Motivation Add composable kernel support on gfx1250. ## Technical Details <!-- Explain the changes along with any relevant GitHub links. --> ## Test Plan <!-- Explain any relevant testing done to verify this PR. --> ## Test Result <!-- Briefly summarize test outcomes. --> ## Submission Checklist - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. --------- Co-authored-by: Qun Lin <qlin@amd.com> Co-authored-by: jialuo12_amdeng <jia.luo@amd.com> Co-authored-by: Andriy Roshchenko <andriy.roshchenko@amd.com> Co-authored-by: hsivasun_amdeng <haresh.sivasuntharampillai@amd.com>
99 lines
4.4 KiB
C++
99 lines
4.4 KiB
C++
// Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
|
|
// SPDX-License-Identifier: MIT
|
|
|
|
#include "gtest/gtest.h"
|
|
#include "ck/ck.hpp"
|
|
#include "profiler/profile_gemm_blockscale_wp_impl.hpp"
|
|
|
|
static ck::index_t param_mask = 0xffff;
|
|
static ck::index_t instance_index = -1;
|
|
|
|
namespace ck {
|
|
namespace test {
|
|
|
|
using Row = ck::tensor_layout::gemm::RowMajor;
|
|
using F32 = float;
|
|
|
|
template <typename Tuple>
|
|
class TestGemmBlockscaleWPCommon : public ::testing::Test
|
|
{
|
|
protected:
|
|
using ALayout = std::tuple_element_t<0, Tuple>;
|
|
using BLayout = std::tuple_element_t<1, Tuple>;
|
|
using CLayout = Row;
|
|
using A0DataType = std::tuple_element_t<2, Tuple>;
|
|
using A1DataType = std::tuple_element_t<3, Tuple>;
|
|
using B0DataType = std::tuple_element_t<4, Tuple>;
|
|
using B1DataType = std::tuple_element_t<5, Tuple>;
|
|
using ComputeDataType = std::tuple_element_t<6, Tuple>;
|
|
using CDataType = std::tuple_element_t<7, Tuple>;
|
|
|
|
public:
|
|
static constexpr bool verify_ = true;
|
|
static constexpr int init_method_ = 1;
|
|
static constexpr bool log_ = false;
|
|
static constexpr bool bench_ = false;
|
|
static constexpr index_t ScaleBlockM = 1;
|
|
static constexpr index_t ScaleBlockN = 128;
|
|
static constexpr index_t ScaleBlockK = 128;
|
|
|
|
void Run(const int M, const int N, const int K, int n_warmup = 1, int n_iter = 10)
|
|
{
|
|
bool all_success = true;
|
|
|
|
int StrideA = std::is_same_v<ALayout, Row> ? K : M;
|
|
int StrideB = std::is_same_v<BLayout, Row> ? N : K;
|
|
int StrideC = std::is_same_v<CLayout, Row> ? N : M;
|
|
|
|
all_success =
|
|
all_success &
|
|
ck::profiler::profile_gemm_blockscale_weightpreshuffle_impl<A0DataType,
|
|
A1DataType,
|
|
B0DataType,
|
|
B1DataType,
|
|
ComputeDataType,
|
|
F32,
|
|
CDataType,
|
|
ScaleBlockM,
|
|
ScaleBlockN,
|
|
ScaleBlockK,
|
|
ALayout,
|
|
BLayout,
|
|
CLayout>(verify_,
|
|
init_method_,
|
|
log_,
|
|
bench_,
|
|
M,
|
|
N,
|
|
K,
|
|
StrideA,
|
|
StrideB,
|
|
StrideC,
|
|
n_warmup,
|
|
n_iter,
|
|
0,
|
|
instance_index);
|
|
|
|
EXPECT_TRUE(all_success);
|
|
}
|
|
};
|
|
|
|
} // namespace test
|
|
} // namespace ck
|
|
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();
|
|
}
|