ROCm / rocprofiler-compute

Advanced Profiling and Analytics for AMD Hardware
https://rocm.docs.amd.com/projects/omniperf/en/latest/
MIT License
135 stars 49 forks source link

None of instruction mix panels contain min/max/avg breakdown across dispatches #154

Closed skyreflectedinmirrors closed 1 year ago

skyreflectedinmirrors commented 1 year ago

E.g.,:

https://github.com/AMDResearch/omniperf/blob/6487ba3853deaf97a91e03c0dc4d8a8109099546/src/omniperf_analyze/configs/gfx90a/1000_compute-unit-instruction-mix.yaml#L24

image

This is just implicitly giving you an average over all selected dispatches, which is pretty different from all the other tables.

skyreflectedinmirrors commented 1 year ago

Same issue exists for the compute pipeline SOL:

https://github.com/AMDResearch/omniperf/blob/ed31b8a988b0fde6a5e11bc949417c82c6db1abc/src/omniperf_analyze/configs/gfx90a/1100_compute-unit-compute-pipeline.yaml#L20

coleramos425 commented 1 year ago

Implemented