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

Replace hipDeviceSynchronize() by hipStreamSynchronize(stream) calls (#1359)

Browse Source
This commit is contained in:
Po Yen Chen
2024-06-26 22:04:52 +08:00
committed by GitHub
parent 0cb2e06ddc
commit a32b1bc647
1 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
include/ck_tile/host/timer.hpp
View File

@@ -27,7 +27,7 @@ struct gpu_timer
CK_TILE_HOST void start(const hipStream_t& s)
{
HIP_CHECK_ERROR(hipDeviceSynchronize());
HIP_CHECK_ERROR(hipStreamSynchronize(s));
HIP_CHECK_ERROR(hipEventRecord(start_evt, s));
}
@@ -51,15 +51,15 @@ struct gpu_timer
struct cpu_timer
{
// torch.utils.benchmark.Timer(), there is a sync inside each timer callback
CK_TILE_HOST void start(const hipStream_t&)
CK_TILE_HOST void start(const hipStream_t& s)
{
HIP_CHECK_ERROR(hipDeviceSynchronize());
HIP_CHECK_ERROR(hipStreamSynchronize(s));
start_tick = std::chrono::high_resolution_clock::now();
}
// torch.utils.benchmark.Timer(), there is a sync inside each timer callback
CK_TILE_HOST void stop(const hipStream_t&)
CK_TILE_HOST void stop(const hipStream_t& s)
{
HIP_CHECK_ERROR(hipDeviceSynchronize());
HIP_CHECK_ERROR(hipStreamSynchronize(s));
stop_tick = std::chrono::high_resolution_clock::now();
}
// return in ms