chore(deps): update bumpalo crate

[overlookmotel]

Latest version of bumpalo includes a couple of performance fixes for String (e.g. fitzgen/bumpalo#229) which may help the parser a little.

[overlookmotel]

Current dependencies on/for this PR:

• main
  • PR chore(deps): update bumpalo crate #2417 👈

This stack of pull requests is managed by Graphite.

[codspeed-hq]

CodSpeed Performance Report

Merging #2417 will not alter performance

_{Comparing 02-15-chore_deps_update_bumpalo_crate (607227a) with main (3ff3495)}

Summary

✅ 27 untouched benchmarks

[overlookmotel]

What the hell? It actually hurts the benchmarks (very slightly), rather than improving them. I'll have to look into this.

[Boshen]

Next time you can use bumpalo = { git = "your git fork url" } and push the branch here before submitting PRs upstream.

[overlookmotel]

This is a weird one. It was definitely an improvement for bumpalo. It just replicates what std's String does, and produces an 80x speed-up for long strings. Question is why is it not having that positive effect in OXC? Probably it's that JS string literals are typically quite short, so it's not hitting the "sweet spot", but I'd like to bore into it a bit more and check that assumption.

Thanks for the tip about git forks. Yes, I'll do that next time.

[Boshen]

It's probably the benchmark not hitting the sweat spots. Let's merge either way.

[overlookmotel]

OK, thanks for merging.

We can probably do better in the parser as building strings now happens in blocks of 32 bytes, so can write each block with 2 x xmm fixed-length writes rather than calls to copy_nonoverlapping.

#2295 and #2409 may both require a bit of an overhaul of string handling, so can look at it then.