mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-12 17:26:00 +00:00
3e6c2610ae
* add intrin_mfma_f64_16x16x4f64 * add example * gemm reference add double data type * chang init data * fix M N PerXdlops * fix ifdef * add comparsion config * add conv fwd example * format log out * change rc matrix egister layout * reorganize example * reorganize example 2 * format,because merge develop * fix call impl adding acc data type * lost ; * add compiler warning * change example tunning parameters * add test for fp64 * add instance * add test/gemm/gemm_fp64.cpp * fix get name issue * remove some tunning parameter * fix conflict * format * use integer value for GEMM test * add acc data type * remove typeid because fp16 * fix streamconfig etc bug from merging develop * format * remove test_gemm_xdl_fp64 * add AccDataType * AccDataType problem Co-authored-by: qinletao <letaoqin@amd.com> Co-authored-by: Chao Liu <chao.liu2@amd.com>
160 lines
6.9 KiB
C++
160 lines
6.9 KiB
C++
#include <algorithm>
|
|
#include <cstdlib>
|
|
#include <half.hpp>
|
|
#include <iostream>
|
|
#include <numeric>
|
|
#include <tuple>
|
|
#include <vector>
|
|
|
|
#include "gemm_util.hpp"
|
|
#include "config.hpp"
|
|
#include "print.hpp"
|
|
#include "device.hpp"
|
|
#include "host_tensor.hpp"
|
|
#include "host_tensor_generator.hpp"
|
|
#include "host_gemm.hpp"
|
|
#include "device_tensor.hpp"
|
|
#include "device_gemm_xdl.hpp"
|
|
#include "device_gemm_xdl_cshuffle.hpp"
|
|
#include "element_wise_operation.hpp"
|
|
#include "reference_gemm.hpp"
|
|
#include "gemm_specialization.hpp"
|
|
|
|
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
|
|
|
|
using DeviceGemmNoOpPtr =
|
|
ck::tensor_operation::device::DeviceGemmPtr<ck::tensor_operation::element_wise::PassThrough,
|
|
ck::tensor_operation::element_wise::PassThrough,
|
|
ck::tensor_operation::element_wise::PassThrough>;
|
|
|
|
namespace ck {
|
|
namespace tensor_operation {
|
|
namespace device {
|
|
namespace device_gemm_instance {
|
|
void add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
|
|
void add_device_gemm_xdl_splitk_f32_f32_f32_km_kn_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_splitk_f32_f32_f32_km_nk_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_splitk_f32_f32_f32_mk_nk_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_splitk_f32_f32_f32_mk_kn_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
|
|
void add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
void add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
|
|
|
|
} // namespace device_gemm_instance
|
|
} // namespace device
|
|
} // namespace tensor_operation
|
|
} // namespace ck
|
|
|
|
int main()
|
|
{
|
|
using ADataType = float;
|
|
using BDataType = float;
|
|
using CDataType = float;
|
|
using AccDataType = float;
|
|
|
|
using RowMajor = ck::tensor_layout::gemm::RowMajor;
|
|
using ColumnMajor = ck::tensor_layout::gemm::ColumnMajor;
|
|
|
|
bool res = true;
|
|
std::vector<DeviceGemmNoOpPtr> gemmPtrs;
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(gemmPtrs);
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_splitk_f32_f32_f32_km_kn_mn_instances(gemmPtrs);
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instances(gemmPtrs);
|
|
|
|
for(auto& gemmPtr : gemmPtrs)
|
|
{
|
|
res &= ck::gemm_util::TestGemm<DeviceGemmNoOpPtr,
|
|
ADataType,
|
|
BDataType,
|
|
CDataType,
|
|
AccDataType,
|
|
ColumnMajor,
|
|
RowMajor,
|
|
RowMajor,
|
|
PassThrough,
|
|
PassThrough,
|
|
PassThrough>{}(gemmPtr);
|
|
}
|
|
|
|
gemmPtrs.clear();
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(gemmPtrs);
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_splitk_f32_f32_f32_km_nk_mn_instances(gemmPtrs);
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(gemmPtrs);
|
|
|
|
for(auto& gemmPtr : gemmPtrs)
|
|
{
|
|
res &= ck::gemm_util::TestGemm<DeviceGemmNoOpPtr,
|
|
ADataType,
|
|
BDataType,
|
|
CDataType,
|
|
AccDataType,
|
|
ColumnMajor,
|
|
ColumnMajor,
|
|
RowMajor,
|
|
PassThrough,
|
|
PassThrough,
|
|
PassThrough>{}(gemmPtr);
|
|
}
|
|
|
|
gemmPtrs.clear();
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(gemmPtrs);
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_splitk_f32_f32_f32_mk_kn_mn_instances(gemmPtrs);
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instances(gemmPtrs);
|
|
|
|
for(auto& gemmPtr : gemmPtrs)
|
|
{
|
|
res &= ck::gemm_util::TestGemm<DeviceGemmNoOpPtr,
|
|
ADataType,
|
|
BDataType,
|
|
CDataType,
|
|
AccDataType,
|
|
RowMajor,
|
|
RowMajor,
|
|
RowMajor,
|
|
PassThrough,
|
|
PassThrough,
|
|
PassThrough>{}(gemmPtr);
|
|
}
|
|
|
|
gemmPtrs.clear();
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(gemmPtrs);
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_splitk_f32_f32_f32_mk_nk_mn_instances(gemmPtrs);
|
|
ck::tensor_operation::device::device_gemm_instance::
|
|
add_device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instances(gemmPtrs);
|
|
|
|
for(auto& gemmPtr : gemmPtrs)
|
|
{
|
|
res &= ck::gemm_util::TestGemm<DeviceGemmNoOpPtr,
|
|
ADataType,
|
|
BDataType,
|
|
CDataType,
|
|
AccDataType,
|
|
RowMajor,
|
|
ColumnMajor,
|
|
RowMajor,
|
|
PassThrough,
|
|
PassThrough,
|
|
PassThrough>{}(gemmPtr);
|
|
}
|
|
|
|
std::cout << "TestGemm ..... " << (res ? "SUCCESS" : "FAILURE") << std::endl;
|
|
return res ? 0 : 1;
|
|
}
|