I also reduced the total number of iterations for some tests as they were order of magnitudes more than necessary from what I can tell. I also made some of these iterative tests parallel via rayon (existing dep) as they weren't utilizing 100% CPU.
The drop in iterations in some of these changes shouldn't have much of an impact on the actual math tests as it's random numbers being tested. We may also want to try a fuzzing crate or something for more deterministic or aggressive targeting rather than rng.
Test Durations Before:
Test Durations After:
(Aarch64 tests are slower due to running on a Medium instance instead of x86's Docker Large)
We should be able to improve build speed a bit more with sscache as well.
I also reduced the total number of iterations for some tests as they were order of magnitudes more than necessary from what I can tell. I also made some of these iterative tests parallel via rayon (existing dep) as they weren't utilizing 100% CPU.
The drop in iterations in some of these changes shouldn't have much of an impact on the actual math tests as it's random numbers being tested. We may also want to try a fuzzing crate or something for more deterministic or aggressive targeting rather than rng.
Test Durations Before:
Test Durations After:
(Aarch64 tests are slower due to running on a Medium instance instead of x86's Docker Large)
We should be able to improve build speed a bit more with sscache as well.