Optimize `try_expand_impl_trait_type` #65293

tmandry · 2019-10-11T00:06:41Z

A lot of time was being spent expanding some large impl Future types in fuchsia. This PR takes the number of types being visited in one expansion from >3 billion to about a thousand, and eliminates the compile time regression in #65147 (in fact, compile times are better than they were before).

Thanks to @Mark-Simulacrum for helping identify the issue and to @matthewjasper for suggesting this change.

Fixes #65147.
r? @matthewjasper,@nikomatsakis

tmandry · 2019-10-11T00:41:31Z

@bors try

bors · 2019-10-11T00:41:48Z

⌛ Trying commit 6493e6b with merge 0d2ff89f766a2f16a09c41ee7c65453980e81908...

bors · 2019-10-11T03:34:40Z

☀️ Try build successful - checks-azure
Build commit: 0d2ff89f766a2f16a09c41ee7c65453980e81908 (0d2ff89f766a2f16a09c41ee7c65453980e81908)

tmandry · 2019-10-11T03:41:01Z

@rust-timer queue

rust-timer · 2019-10-11T03:41:02Z

Awaiting bors try build completion

tmandry · 2019-10-11T03:46:32Z

@rust-timer build 0d2ff89f766a2f16a09c41ee7c65453980e81908

rust-timer · 2019-10-11T03:46:33Z

Queued 0d2ff89f766a2f16a09c41ee7c65453980e81908 with parent 898f36c, future comparison URL.

Centril · 2019-10-11T06:49:13Z

src/librustc/ty/util.rs

@@ -697,6 +697,9 @@ impl<'tcx> TyCtxt<'tcx> {
            // that type, and when we finish expanding that type we remove the
            // its DefId.
            seen_opaque_tys: FxHashSet<DefId>,
+            // Cache of all expansions we've seen so far. This is a critical
+            // optimization for some large types produced by async fn trees.


Suggested change

// optimization for some large types produced by async fn trees.

// optimization for some large types produced by `async fn` trees.

Centril · 2019-10-11T06:50:07Z

src/librustc/ty/util.rs

-                    let concrete_ty = generic_ty.subst(self.tcx, substs);
-                    let expanded_ty = self.fold_ty(concrete_ty);
-                    self.seen_opaque_tys.remove(&def_id);
+                    let expanded_ty = match self.expanded_cache.get(&def_id) {


Why not use the .entry API?

Yeah! Entry! Entry! I ❤️ Entry!

Oh, I think I see why not -- the self.fold_ty call would error out with entry, wouldn't it?

Yeah, the fact that the closure would need &mut self prevents me from using entry here :/

src/test/ui/impl-trait/recursive-impl-trait-type--through-non-recursize.stderr

nikomatsakis · 2019-10-11T17:03:19Z

@tmandry I'm not sure if we have a suitable test already, but if not, can you add something to perf that reflects this compile time gap?

nikomatsakis

Presuming this helps, and if we add a perf regression test, seems good to me! Nice detective work y'all.

rust-timer · 2019-10-11T20:16:32Z

Finished benchmarking try commit 0d2ff89f766a2f16a09c41ee7c65453980e81908, comparison URL.

tmandry · 2019-10-11T22:22:25Z

@tmandry I'm not sure if we have a suitable test already, but if not, can you add something to perf that reflects this compile time gap?

Will do. I was hoping something would turn up in the existing test suite, but it doesn't look that way :)

matthewjasper

I think the following test case shows the problem here:

#![type_length_limit="10000000"]
#![allow(unused)]

macro_rules! mk_async_fn {
    ($f:ident $g:ident) => {
        async fn $g() -> i32 {
            $f().await;
            $f().await;
            $f().await
        }
    }
}

async fn a() -> i32 { 1 }

mk_async_fn!(a b);
mk_async_fn!(b c);
mk_async_fn!(c d);
mk_async_fn!(d e);
mk_async_fn!(e f);
mk_async_fn!(f g);
mk_async_fn!(g h);
mk_async_fn!(h i);
mk_async_fn!(i j);

src/librustc/ty/util.rs

tmandry · 2019-10-14T22:50:45Z

@bors try

bors · 2019-10-14T22:50:55Z

⌛ Trying commit 6493e6b with merge 0c36cd3...

@matthewjasper

Optimize `try_expand_impl_trait_type` A lot of time was being spent expanding some large `impl Future` types in fuchsia. This PR takes the number of types being visited in one expansion from >3 billion to about a thousand, and eliminates the compile time regression in #65147 (in fact, compile times are better than they were before). Thanks to @matthewjasper for suggesting this change. Fixes #65147. r? @matthewjasper,@nikomatsakis

tmandry · 2019-10-15T01:18:17Z

@matthewjasper thanks for supplying the benchmark!

This should be ready, pending another timer run once rust-lang/rustc-perf#508 is merged.

bors · 2019-10-15T01:56:00Z

☀️ Try build successful - checks-azure
Build commit: 0c36cd3 (0c36cd32207594f45af4c9c92294f769113808b6)

tmandry · 2019-10-15T19:40:47Z

@bors r=matthewjasper

bors · 2019-10-15T19:40:48Z

📌 Commit 802554f has been approved by matthewjasper

@Mark-Simulacrum

…sper Optimize `try_expand_impl_trait_type` A lot of time was being spent expanding some large `impl Future` types in fuchsia. This PR takes the number of types being visited in one expansion from >3 billion to about a thousand, and eliminates the compile time regression in rust-lang#65147 (in fact, compile times are better than they were before). Thanks to @Mark-Simulacrum for helping identify the issue and to @matthewjasper for suggesting this change. Fixes rust-lang#65147. r? @matthewjasper,@nikomatsakis

@ghost

Rollup of 14 pull requests Successful merges: - #64603 (Reducing spurious unused lifetime warnings.) - #64623 (Remove last uses of gensyms) - #65235 (don't assume we can *always* find a return type hint in async fn) - #65242 (Fix suggestion to constrain trait for method to be found) - #65265 (Cleanup librustc mir err codes) - #65293 (Optimize `try_expand_impl_trait_type`) - #65307 (Try fix incorrect "explicit lifetime name needed") - #65308 (Add long error explanation for E0574) - #65353 (save-analysis: Don't ICE when resolving qualified type paths in struct members) - #65389 (Return `false` from `needs_drop` for all zero-sized arrays.) - #65402 (Add troubleshooting section to PGO chapter in rustc book.) - #65425 (Optimize `BitIter`) - #65438 (Organize `never_type` tests) - #65444 (Implement AsRef<[T]> for List<T>) Failed merges: - #65390 (Add long error explanation for E0576) r? @ghost

pnkfelix · 2019-10-17T14:24:45Z

discussed at T-compiler meeting. Accepted for beta-backport.

[beta] backport rollup This includes a bunch of PRs: * Fix redundant semicolon lint interaction with proc macro attributes #64387 * Upgrade async/await to "used" keywords. #64875 * syntax: fix dropping of attribute on first param of non-method assocated fn #64894 * async/await: improve not-send errors #64895 * Silence unreachable code lint from await desugaring #64930 * Always mark rust and rust-call abi's as unwind #65020 * Account for macro invocation in `let mut $pat` diagnostic. #65123 * Ensure that associated `async fn`s have unique fresh param names #65142 * Add troubleshooting section to PGO chapter in rustc book. #65402 * Upgrade GCC to 8.3.0, glibc to 1.17.0 and crosstool-ng to 1.24.0 for dist-armv7-linux #65302 * Optimize `try_expand_impl_trait_type` #65293 * use precalculated dominators in explain_borrow #65172 * Fix ICE #64964 #64989

[beta] backport rollup This includes a bunch of PRs: * Fix redundant semicolon lint interaction with proc macro attributes #64387 * Upgrade async/await to "used" keywords. #64875 * syntax: fix dropping of attribute on first param of non-method assocated fn #64894 * async/await: improve not-send errors #64895 * Silence unreachable code lint from await desugaring #64930 * Always mark rust and rust-call abi's as unwind #65020 * Account for macro invocation in `let mut $pat` diagnostic. #65123 * Ensure that associated `async fn`s have unique fresh param names #65142 * Add troubleshooting section to PGO chapter in rustc book. #65402 * Upgrade GCC to 8.3.0, glibc to 1.17.0 and crosstool-ng to 1.24.0 for dist-armv7-linux #65302 * Optimize `try_expand_impl_trait_type` #65293 * use precalculated dominators in explain_borrow #65172 * Fix ICE #64964 #64989 * [beta] Revert "Auto merge of #62948 - matklad:failable-file-loading, r=petro… #65273 * save-analysis: Don't ICE when resolving qualified type paths in struct members #65353 * save-analysis: Nest tables when processing impl block definitions #65511 * Avoid ICE when checking `Destination` of `break` inside a closure #65518 * Avoid ICE when adjusting bad self ty #65755 * workaround msys2 bug #65762

rust-highfive assigned matthewjasper Oct 11, 2019

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 11, 2019

Centril reviewed Oct 11, 2019

View reviewed changes

src/test/ui/impl-trait/recursive-impl-trait-type--through-non-recursize.stderr Outdated Show resolved Hide resolved

nikomatsakis approved these changes Oct 11, 2019

View reviewed changes

matthewjasper requested changes Oct 12, 2019

View reviewed changes

src/librustc/ty/util.rs Outdated Show resolved Hide resolved

src/librustc/ty/util.rs Outdated Show resolved Hide resolved

matthewjasper added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 12, 2019

tmandry mentioned this pull request Oct 14, 2019

Add await-call-tree benchmark rust-lang/rustc-perf#508

Merged

tmandry added 2 commits October 14, 2019 15:48

Add expanded type cache to OpaqueTypeExpander

fe09bb5

Only expand types that contain projections

802554f

tmandry added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Oct 14, 2019

tmandry force-pushed the turbo-expander branch from 6493e6b to 802554f Compare October 15, 2019 19:17

matthewjasper approved these changes Oct 15, 2019

View reviewed changes

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 15, 2019

tmandry added beta-nominated Nominated for backporting to the compiler in the beta channel. I-nominated T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Oct 15, 2019

tmandry mentioned this pull request Oct 15, 2019

Rollup of 14 pull requests #65454

Merged

bors merged commit 802554f into rust-lang:master Oct 16, 2019

pnkfelix added the beta-accepted Accepted for backporting to the compiler in the beta channel. label Oct 17, 2019

Mark-Simulacrum removed beta-nominated Nominated for backporting to the compiler in the beta channel. I-nominated labels Oct 22, 2019

Mark-Simulacrum mentioned this pull request Oct 22, 2019

[beta] backport rollup #65708

Merged

ibraheemdev mentioned this pull request Jul 25, 2021

Exponential compile time with recursive opaque type #87450

Closed

	// optimization for some large types produced by async fn trees.
	// optimization for some large types produced by `async fn` trees.

Optimize try_expand_impl_trait_type #65293

Optimize try_expand_impl_trait_type #65293

Uh oh!

Conversation

tmandry commented Oct 11, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tmandry commented Oct 11, 2019

Uh oh!

bors commented Oct 11, 2019

Uh oh!

bors commented Oct 11, 2019

Uh oh!

tmandry commented Oct 11, 2019

Uh oh!

rust-timer commented Oct 11, 2019

Uh oh!

tmandry commented Oct 11, 2019

Uh oh!

rust-timer commented Oct 11, 2019

Uh oh!

Centril Oct 11, 2019

Choose a reason for hiding this comment

Uh oh!

Centril Oct 11, 2019

Choose a reason for hiding this comment

Uh oh!

nikomatsakis Oct 11, 2019

Choose a reason for hiding this comment

Uh oh!

nikomatsakis Oct 11, 2019

Choose a reason for hiding this comment

Uh oh!

tmandry Oct 14, 2019

Choose a reason for hiding this comment

Uh oh!

Uh oh!

nikomatsakis commented Oct 11, 2019

Uh oh!

nikomatsakis left a comment

Choose a reason for hiding this comment

Uh oh!

rust-timer commented Oct 11, 2019

Uh oh!

tmandry commented Oct 11, 2019

Uh oh!

matthewjasper left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

tmandry commented Oct 14, 2019

Uh oh!

bors commented Oct 14, 2019

Uh oh!

tmandry commented Oct 15, 2019

Uh oh!

bors commented Oct 15, 2019

Uh oh!

tmandry commented Oct 15, 2019

Uh oh!

bors commented Oct 15, 2019

Uh oh!

pnkfelix commented Oct 17, 2019

Uh oh!

Uh oh!

Optimize `try_expand_impl_trait_type` #65293

Optimize `try_expand_impl_trait_type` #65293

tmandry commented Oct 11, 2019 •

edited

Loading