Performance regression in 1.72 related to normalization of opaques with late-bound vars

[kornelski]

Upgrade from 1.71 to 1.72 has made compilation time of my async-heavy actix server 350 times slower (from under 5s to 30 minutes, on a 32GB M1 Max CPU).

I've bisected it to commit a20a04e (#113108).

The slowdown is still in the latest nightly (1.74.0 69e97df 2023-08-26). 1.72 and the nightly spend 30% of all compilation time in Interners::intern_ty, mainly called by <rustc_middle[b15c2eca62e7285b]::ty::Ty as rustc_type_ir[f43ddb3fb00a443c]::fold::TypeSuperFoldable<rustc_middle[b15c2eca62e7285b]::ty::context::TyCtxt>>::try_super_fold_with::<rustc_middle[b15c2eca62e7285b]::ty::generic_args::ArgFolder> and <rustc_middle[b15c2eca62e7285b]::ty::Ty as rustc_type_ir[f43ddb3fb00a443c]::fold::TypeSuperFoldable<rustc_middle[b15c2eca62e7285b]::ty::context::TyCtxt>>::try_super_fold_with::<rustc_trait_selection[961e83e42176765a]::traits::project::AssocTypeNormalizer>.

The reduced code is in: https://gitlab.com/lib.rs/main/-/commits/reproducer

outdated info

Apologies for a big spaghetti reproducer. Let me know if it'd be useful to reduce it.

git clone --recursive https://gitlab.com/lib.rs/main
cd main
git reset --hard 81c612b
cargo build -p crates-server

I suspect the culprit is this function, which is used in several places in server/src/main.rs:

async fn with_file_cache<F: Send>(state: &AServerState, cache_file_name: &str, cache_time: u32, allow_stale: bool, generate: F) -> Result<Rendered, anyhow::Error>
    where F: Future<Output=Result<Rendered, anyhow::Error>> + 'static {

@rustbot modify labels: +regression-from-stable-to-stable -regression-untriaged

[compiler-errors]

Yeah, I'd appreciate if you could minimize it. @Dirbaio helped me with a similar regression that was fixed in #114948, so I wonder if there's some other missing normalization that needs to be added somewhere...

[compiler-errors]

In the mean time, I'll try to look at the flamegraphs for the code you shared...

[compiler-errors]

@kornelski: Is there another step I need to do in this build? 😅

error: couldn't read front_end/src/../../style/public/critical.css: No such file or directory (os error 2)
   --> front_end/src/front_end.rs:101:53
    |
101 |         let data = self.critical_css_data.unwrap_or(include_str!("../../style/public/critical.css"));
    |                                                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

[kornelski]

Thank you for checking it out — sorry about the build error. I though I removed all of these. CSS is irrelevant here, can be an empty string.

https://gitlab.com/lib.rs/main/-/commits/reproducer

[compiler-errors]

I'm finding several main sources of slowness here:

1. Monomorphization causing us to repeatedly normalize opaques that show up in MIR bodies -- enable RevealAll pass always #115299 causes the compile time to go from 28 minutes to 11 minutes, for example.
2. We're repeatedly checking FnOnce obligations for the FnDef¹ types of async fns. This leads us to repeatedly compute the signatures of these async fns and then normalize their output futures over and over again.

Footnotes

1. The ZST corresponding to each function item. ↩

[apiraino]

WG-prioritization assigning priority (Zulip discussion).

@rustbot label -I-prioritize +P-high