Given that you are making a bunch of changes and performance is important, have you considered adding benching to CI? I feel like I have seen Rust projects that have good performance regression detection in CI, but I might need to do a bit of searching.

This would be nice because performance would be repeatable (hopefully!) unlike

if I have singleplayer Minecraft open on my machine at the same time I'm running bench_players, the performance actually improves.

#323

@andrewgazelka I'm open to the idea, but it seems like a lot of work to do it correctly. For starters we would need dedicated CI runners and a decent suite of benches. Definitely something to think about after the rewrite is finished.

The criterion crate sounds like it may be useful here, unless ya'll have your own favorite (micro)benchmarking tools.

@andrewgazelka I'm open to the idea, but it seems like a lot of work to do it correctly. For starters we would need dedicated CI runners and a decent suite of benches. Definitely something to think about after the rewrite is finished.

That makes sense. Why would you need dedicated CI runners, though? Do you think performance would not be predictable using the default CI runners, which run in a VM on the same machine as other runners?

If consistent environments are available (e.g. on your own machine), it may be worth it to use a microbenchmarking crate like criterion to check for performance regressions.

Do you think performance would not be predictable using the default CI runners, which run in a VM on the same machine as other runners?

That's right, it's too noisy from what I've heard.

@A-z-z-y I'm quite fond of divan nowadays :)

The crates glam and nalgebra are both well-maintained and highly optimized. There's a benchmark that compares a bunch of popular lin alg crates but it's over 3 years old.

Closing this since I'm not happy with the direction this was going and it's too much to do in one PR anyway.