cutlass [QST] Benchmark script

Hi, there is a performance benchmark in README.md, but there is no link to the code to reproduce it. Can you please point me to the source code for this benchmark?

May 23 '25 07:05 JohnAlphaIII

we just use the cutlass profiler and pick the highest perf kernel out of the default kernel set of the profiler for each layout and dtype combo. relevant hardware details etc can be found in the plots themselves.

May 23 '25 13:05 thakkarV

Thanks @thakkarV . Do you mean you were using this script? https://github.com/NVIDIA/cutlass/blob/main/tools/profiler/src/gemm_operation_profiler.cu

May 23 '25 17:05 JohnAlphaIII

https://docs.nvidia.com/cutlass/media/docs/cpp/profiler.html

May 25 '25 19:05 thakkarV

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

Jun 26 '25 10:06 github-actions[bot]