ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_Tile] Updating gpu timer when doing flush cache (#2593)

Browse Source 
* Missed updating function names in example

* updating timer

* code cleanup

* addressing review comments

* updating tile_engine code

* addressing review comments
This commit is contained in:
Khushbu Agarwal
2025-07-31 16:43:33 -07:00
committed by GitHub
parent 546ef78d1d
commit 88d72178d6
11 changed files with 54 additions and 139 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tile_engine/ops/gemm/benchmark_gemm.cpp
View File

@@ -34,8 +34,7 @@ void benchmark_gemm(const ck_tile::ArgParser& arg_parser)
arg_parser.get_bool("log"),
arg_parser.get_str("csv_filename"),
arg_parser.get_bool("flush_cache"),
arg_parser.get_int("rotating_count"),
arg_parser.get_int("bench_time")};
arg_parser.get_int("rotating_count")};
auto& profiler = GemmProfiler::instance(setting);

1
tile_engine/ops/gemm/benchmark_gemm.hpp
View File

@@ -125,7 +125,6 @@ struct Setting
std::string csv_filename_;
bool flush_cache_;
int rotating_count_;
int bench_time_ms_;
};
inline std::string get_rocm_version()

1
tile_engine/ops/gemm/gemm_host_api.hpp
View File

@@ -110,7 +110,6 @@ inline auto create_args(int argc, char* argv[])
"To flush cache, possible values are true or false. "
"Default is false.")
.insert("rotating_count", "5", "number of iterations to rotate the cache. default is 5.")
.insert("bench_time", "0", "benchmark time in ms. default is 0 ms.")
.insert("metric",
"0",
"Metric with which to measure kernel performance. Set to 0 for latency, 1 for "

3
tile_engine/ops/gemm/gemm_profiler.hpp
View File

@@ -131,8 +131,7 @@ class GemmProfiler
setting_.n_repeat_,
setting_.is_gpu_timer_,
setting_.flush_cache_,
setting_.rotating_count_,
setting_.bench_time_ms_});
setting_.rotating_count_});
process_result(gemm_problem,
c_m_n_dev_buf,
c_m_n_host_result,