ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-13 01:36:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add gridwise GEMM pipeline (#89)

Browse Source 
* clean up

* add mutilple thread scratch to ThreadwiseTensorSliceTransfer_v3r1

* add 2 stage prefetch

* add more sanity check into transform_tensor_descriptor

* tweak

* enabling 2 stage prefetch to exsiting gridwise gemm; tweak

* enabling 2 stage prefetch to exsiting gridwise gemm

* move gridwise gemm pipeline in class; clean up

* add some irregular tile size

* update CalculateHasMainK0BlockLoop for multi-stage-prefetch

* refactor gridwise gemm pipeline class
This commit is contained in:
Chao Liu
2022-02-23 17:23:49 -06:00
committed by GitHub
parent 756a761727
commit 22d438ae9e
18 changed files with 864 additions and 1694 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
profiler/include/profile_gemm_impl.hpp
View File

@@ -1,4 +1,5 @@
#pragma once
#include <iomanip>
#include "config.hpp"
#include "device.hpp"
#include "host_tensor.hpp"
@@ -30,6 +31,9 @@ void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(std::vector<De
void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
void add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_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_f32_f32_f32_mk_nk_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
void add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(std::vector<DeviceGemmNoOpPtr>&);
@@ -225,6 +229,9 @@ void profile_gemm_impl(int do_verification,
ck::tensor_operation::device::device_gemm_instance::
add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(gemm_ptrs);
ck::tensor_operation::device::device_gemm_instance::
add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(gemm_ptrs);
}
else if constexpr(is_same<ALayout, tensor_layout::gemm::ColumnMajor>::value &&
is_same<BLayout, tensor_layout::gemm::RowMajor>::value &&
@@ -293,8 +300,8 @@ void profile_gemm_impl(int do_verification,
float gb_per_sec = num_btype / 1.E6 / ave_time;
std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec
<< " GB/s, " << gemm_name << std::endl;
std::cout << "Perf: " << std::setw(10) << ave_time << " ms, " << tflops << " TFlops, "
<< gb_per_sec << " GB/s, " << gemm_name << std::endl;
if(tflops > best_tflops)
{