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

Fix roofline calculation for single dispatch #97

Closed coleramos425 closed 1 year ago

coleramos425 commented 1 year ago

As pointed out my an Omniperf user. An issue with roofline calculation arrises in a workload with a single dispatch. https://github.com/AMDResearch/omniperf/blob/0705db024c184bb8b99e64eea36b1cf36c062e0a/src/omniperf_analyze/utils/roofline_calc.py#L211-L221 Such that, a single dispatch is overlooked in for-loop logic. This code requires adjustment.

coleramos425 commented 1 year ago

Fix implemented and merged into main. Closing ticket.