timemory or python-papi

[pwm1234-sri]

Hello! It has been a couple of years since I have looked a timemory. And it looks like the pace of development on timemory has slowed down - the last commit on the develop branch was March 2022. I am curious - is that because timemory has matured and there is little left to be done or because interest has moved on to something else?

I am back re-visiting timemory because I am working on a new project for which performance profiling will be important. I need to keep track of both wall clock time and computational complexity (at least FLOPS). I came across Counting FLOPS and other CPU counters in Python | B. Nikolic Software and Computing Blog and it made me think - should I use python-papi, which looks to be quite simple for estimating FLOPS, or should I re-invest in timemory? I appreciate any feedback.

Thanks,
Phil

[jrmadsen]

Hi Phil, there has been plenty of activity in the last year or two but, for various reasons, I have been holding off on merging all the updates in the omnitrace branch for far too long now. It is being used extensively for a lot of functionality in AMDResearch/omnitrace. Since this is the second question about its activity in the last week or so, I’ll probably give up hope of breaking the updates in the omnitrace branch into smaller pieces and merge it in bulk into develop and just bump the major version number (not a whole lot has fundamentally changed though).

[pwm1234-sri]

Thanks, I appreciate your response. I will add timemory to my development tools and see if I can use it. I will let you know if I have problems.