ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-17 00:58:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

update preprocess number for mi308; bring back printout in ckprofiler

Browse Source
This commit is contained in:
aska-0096
2024-10-25 04:29:34 +00:00
parent 47294b4b22
commit e8c19535f7
2 changed files with 7 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
include/ck/host_utility/flush_cache.hpp
View File

@@ -351,7 +351,10 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
}
#else
// return total_time / nrepeat;
return (total_time - 0.01 * nrepeat) / nrepeat;
hipDeviceProp_t deviceProps;
hip_check_error(hipGetDeviceProperties(&deviceProps, 0));
float preprocess_offset = deviceProps.multiProcessorCount==80? 0.005 : 0.01;
return (total_time - preprocess_offset * nrepeat) / nrepeat;
#endif
}
else

6
profiler/include/profiler/profile_gemm_multiply_multiply_impl.hpp
View File

@@ -267,9 +267,9 @@ bool profile_gemm_multiply_multiply_impl(int do_verification,
float gb_per_sec = num_btype / 1.E6 / ave_time;
// std::cout << "Perf: " << std::setw(10) << ave_time << " ms, " << tflops
// << " TFlops, " << gb_per_sec << " GB/s, " << op_name << ", KBatch "
// << kbatch_curr << std::endl;
std::cout << "Perf: " << std::setw(10) << ave_time << " ms, " << tflops
<< " TFlops, " << gb_per_sec << " GB/s, " << op_name << ", KBatch "
<< kbatch_curr << std::endl;
#if defined CK_ENABLE_FP8 || defined CK_ENABLE_INT8
// set softer tolerances for fp8