Thanks for the detailed analysis, especially regarding the weird behavior of nvtx. I found the comparison of profiling approaches quite helpful.
Are you aware of any updates to your the profiling methods regarding their precision since the time of your analyses?
I have not tested it since then. I think there were multiple projects that tried to improve profilers for PyTorch. I don’t know how they went.
Thanks for the detailed analysis, especially regarding the weird behavior of nvtx. I found the comparison of profiling approaches quite helpful.
Are you aware of any updates to your the profiling methods regarding their precision since the time of your analyses?
I have not tested it since then. I think there were multiple projects that tried to improve profilers for PyTorch. I don’t know how they went.