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

updating timer

Browse Source
This commit is contained in:
khuagarw
2025-07-31 00:22:53 +00:00
parent 343ed414c3
commit 221f08d10a
3 changed files with 60 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
example/ck_tile/03_gemm/gemm_utils.hpp
View File

@@ -458,6 +458,8 @@ auto create_args(int argc, char* argv[])
.insert("split_k", "1", "splitK value")
.insert("init", "0", "0:random, 1:linear, 2:constant(1)")
.insert("persistent", "0", "0:non-persistent, 1:persistent")
.insert("flush_cache", "true", "flush cache before running the kernel, defaults to true")
.insert("rotating_count", "1", "rotating count, defaults to 1")
.insert("bench_time_ms", "0", "benchmark time in ms, defaults to 0 ms");
bool result = arg_parser.parse(argc, argv);

34
example/ck_tile/03_gemm/run_gemm_example.inc
View File

@@ -184,6 +184,8 @@ float invoke_gemm(ck_tile::DeviceMem& a_m_k_dev_buf,
int n_warmup,
int n_repeat,
bool persistent,
bool flush_cache,
int rotating_count,
int bench_time_ms)
{
ck_tile::GemmHostArgs args = {a_m_k_dev_buf.GetDeviceBuffer(),
@@ -211,10 +213,16 @@ float invoke_gemm(ck_tile::DeviceMem& a_m_k_dev_buf,
DsLayout,
CLayout,
true,
CDEElementWise>(
args,
ck_tile::stream_config{
nullptr, true, 1, n_warmup, n_repeat, true, true, 50, bench_time_ms});
CDEElementWise>(args,
ck_tile::stream_config{nullptr,
true,
1,
n_warmup,
n_repeat,
true,
flush_cache,
rotating_count,
bench_time_ms});
}
else
{
@@ -229,10 +237,16 @@ float invoke_gemm(ck_tile::DeviceMem& a_m_k_dev_buf,
DsLayout,
CLayout,
false,
CDEElementWise>(
args,
ck_tile::stream_config{
nullptr, true, 1, n_warmup, n_repeat, true, true, 50, bench_time_ms});
CDEElementWise>(args,
ck_tile::stream_config{nullptr,
true,
1,
n_warmup,
n_repeat,
true,
flush_cache,
rotating_count,
bench_time_ms});
}
std::size_t flop = std::size_t(2) * M * N * K;
@@ -303,6 +317,8 @@ int run_gemm_example_with_layouts(int argc,
int n_repeat = arg_parser.get_int("repeat");
ck_tile::index_t init_method = arg_parser.get_int("init");
bool persistent = arg_parser.get_int("persistent");
bool flush_cache = arg_parser.get_bool("flush_cache");
int rotating_count = arg_parser.get_int("rotating_count");
int bench_time_ms = arg_parser.get_int("bench_time_ms");
const bool preshuffle = GemmConfig::Preshuffle;
@@ -422,6 +438,8 @@ int run_gemm_example_with_layouts(int argc,
n_warmup,
n_repeat,
persistent,
flush_cache,
rotating_count,
bench_time_ms);
c_m_n_dev_buf.FromDevice(c_m_n_dev_result.data());