item-bodies/item-types timings regressed for issue-32062-equality-relations

[nagisa]

See this and this.

[nagisa]

cc @eddyb @nikomatsakis

[jonas-schievink]

cc @Marwes

[Marwes]

While testing #33816 on my projects I figured I'd run through this example on that version of rustc as well and it seems it doesn't seem to give this slowdown and given that #33816 seems to be rebased on top of master after this slowdown occurred it might have been fixed by that PR (if only by chance?). I may take a look at this sometime this weekend if I get the time though as it still seems unfortunate. (It would be really nice if perf.rust-lang.org could display the commit hash of each run it does. That would make it much easier to pin down which version introduced a regression).

[nikomatsakis]

Probably the same problem is causing a regr in jld-day15-parser

[nikomatsakis]

triage: P-high

[nikomatsakis]

It'd be great to do some sort of bisection to try to narrow down what is causing this.

[eddyb]

@nikomatsakis Doesn't look like it, the regression is in translation for jld-day15-parser.

[arielb1]

The problematic PR here is 75e23e1 (#33137). Seems logical - that probably broke the caching

[eddyb]

@arielb1 So the solution is to move forward with #33816?

[arielb1]

That sounds like it.

[arielb1]

This is fixed as #33816 landed.