ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 13:11:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

Universal gemm flush cache (#1251)

Browse Source 
* add flush cache to device op

* add flush cache parameter to ckProfiler

* change calculate size a and b method

* chang evaluation time method foro AVERAGE to MEDIAN

* format code

* adjust some code

* fix core dumped

* remove loop call flush icache in kernel

* remove loop(outer) call flush icache

---------

Co-authored-by: letaoqin <letaoqin@amd.com>
This commit is contained in:
ltqin
2024-04-26 04:07:14 +08:00
committed by GitHub
parent b1f8ae379b
commit f448d179b7
6 changed files with 331 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

50
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle_v3.hpp
View File

@@ -15,6 +15,7 @@
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/host_utility/flush_cache.hpp"
namespace ck {
namespace tensor_operation {
@@ -151,14 +152,49 @@ struct DeviceGemm_Xdl_CShuffleV3 : public DeviceGemmV2<ALayout,
const bool has_main_k_block_loop = GridwiseGemm::CalculateHasMainKBlockLoop(K_split);
const auto Run = [&](const auto& kernel) {
if(arg.KBatch > 1)
hipGetErrorString(hipMemsetAsync(arg.p_c_grid,
0,
arg.M * arg.N * sizeof(CDataType),
stream_config.stream_id_));
if(stream_config.flush_cache)
{
Argument arg_ = arg;
ck::utility::RotatingMemWrapper<Argument> rotating_mem(
arg_,
stream_config.rotating_count,
arg_.M * arg_.K * sizeof(ADataType),
arg_.K * arg_.N * sizeof(BDataType));
rotating_mem.Print();
ave_time = launch_and_time_kernel(
stream_config, kernel, dim3(gdx, gdy, gdz), dim3(BlockSize), 0, arg);
auto run_flush_cache = [&]() {
// flush icache
ck::utility::flush_icache();
// rotating mem
rotating_mem.Next();
// clear c mem
if(arg_.KBatch > 1)
hipGetErrorString(hipMemsetAsync(arg_.p_c_grid,
0,
arg_.M * arg_.N * sizeof(CDataType),
stream_config.stream_id_));
};
ave_time = ck::utility::launch_and_time_kernel_with_preprocess<false>(
stream_config,
run_flush_cache,
kernel,
dim3(gdx, gdy, gdz),
dim3(BlockSize),
0,
arg_);
}
else
{
if(arg.KBatch > 1)
hipGetErrorString(hipMemsetAsync(arg.p_c_grid,
0,
arg.M * arg.N * sizeof(CDataType),
stream_config.stream_id_));
ave_time = launch_and_time_kernel(
stream_config, kernel, dim3(gdx, gdy, gdz), dim3(BlockSize), 0, arg);
}
};
constexpr index_t minimum_occupancy =