ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-02 21:27:45 +00:00
Code Issues Packages Projects Releases Wiki Activity

update in profiler

Browse Source
This commit is contained in:
ozturkosu
2025-06-13 05:49:43 -04:00
parent 9403f38a5c
commit 3fa44df1c2
1 changed files with 13 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
profiler/include/profiler/profile_gemm_universal_streamk_impl.hpp
View File

@@ -250,8 +250,19 @@ bool profile_gemm_universal_streamk_impl(int do_verification,
float gb_per_sec = num_btype / 1.E6 / ave_time;
const auto actual_launch_grid_dims = argument_ptr->GetLaunchGridDims();
// const auto actual_launch_grid_dims = argument_ptr->GetLaunchGridDims();
const auto* typed_argument_ptr = dynamic_cast<const GridwiseGemm_xdl_cshuffle_streamk_v3<...>::Argument*>(argument_ptr);
if (typed_argument_ptr)
{
const auto actual_launch_grid_dims = typed_argument_ptr->GetLaunchGridDims();
std::cout << "Actual Grid Dimensions: " << actual_launch_grid_dims.x << "x"
<< actual_launch_grid_dims.y << "x" << actual_launch_grid_dims.z << std::endl;
}
else
{
std::cerr << "Error: Failed to cast argument_ptr to the correct type." << std::endl;
}
std::cout << "Perf: " << std::setw(10) << ave_time << " ms, " << tflops
<< " TFlops, " << gb_per_sec << " GB/s, " << op_name << ", Grid_size "
<< actual_launch_grid_dims.x // Use the x-dimension of the actual launch grid