Jacob-Stevens-Haas
commented
3 months ago Ok, it looks like the basic benchmarking is working. You can see that the commit with an intentional performance regression failed CI.
Two things I'd like to add:
- [x] test memory useage for finite difference
- ~asv publish artifact so that it's possible to browse the results of the benchmarking action (It'll look like pandas, but with only the merge-base and PR as the two commits)~ This'll have to wait for a later PR if it's desired.
I'd like to just leave this with a test for finite difference as a template later expansion and improvement. I think the PR is ready to go now
Hey Andy, I've added some logging to an experiment and found that some derivative methods can be slow. This isn't a surprise, but the enterprise solution is to add benchmarking to CI so that we can start quantifying improvements. This uses Airspeed Velocity (
asv), the same benchmarking package as pandas.Like most PRs that deal with CI, this one may have a variety of commits to try to get it working. I don't know yet how to feed asv previous benchmark results so that it can know performance improvements/regressions.