Clarify main code paths in exhaustiveness checking #78430

Nadrieril · 2020-10-27T03:31:47Z

This PR massively clarifies the main code paths of exhaustiveness checking, by using the Constructor enum to a fuller extent. I've been itching to write it for more than a year, but the complexity of matching consts had prevented me. Behold a massive simplification :D.
This in particular removes a fair amount of duplication between various parts, localizes code into methods of relevant types when applicable, makes some implicit assumptions explicit, and overall improves legibility a lot (or so I hope). Additionally, after my changes, undoing #76918 turned out to be a noticeable perf gain.

As usual I tried my best to make the commits self-contained and easy to follow. I've also tried to keep the code well-commented, but I tend to forget how complex this file is; I'm happy to clarify things as needed.
My measurements show good perf improvements on the two match-heavy benchmarks (-18.0% on unicode_normalization-check! :D); I'd like a perf run to check the overall impact.

r? @varkor
@rustbot modify labels: +A-exhaustiveness-checking

First is checking for constructor overlap, second is extracting the resulting fields.

Since the constructor is recomputed a lot, caching is worth it.

This is even a perf improvement on the match-heavy benchmarks.

This only happens in a slow (diagnostics) path, so the code clarity gain is worth it.

Also removes the ugly caching that was introduced in rust-lang#76918. It was bolted on without deeper knowledge of the workings of the algorithm. This commit manages to be more performant without any of the complexity. It should be better on representative workloads too.

The test change is because we used to treat `&str` like other `&T`s, ie as having a single constructor. That's not quite true though since we consider `&str` constants as atomic instead of refs to `str` constants.

After splitting, subtraction becomes much simpler

jyn514 · 2020-10-27T14:43:40Z

@bors try @rust-timer queue

rust-timer · 2020-10-27T14:43:42Z

Awaiting bors try build completion

bors · 2020-10-27T14:44:01Z

⌛ Trying commit 766ab78 with merge 5fa81289ac2b01e09c5de71dc65a406e9e69d839...

bors · 2020-10-27T15:27:26Z

☀️ Try build successful - checks-actions
Build commit: 5fa81289ac2b01e09c5de71dc65a406e9e69d839 (5fa81289ac2b01e09c5de71dc65a406e9e69d839)

rust-timer · 2020-10-27T15:27:28Z

Queued 5fa81289ac2b01e09c5de71dc65a406e9e69d839 with parent 56d288f, future comparison URL.

rust-timer · 2020-10-27T23:30:37Z

Finished benchmarking try commit (5fa81289ac2b01e09c5de71dc65a406e9e69d839): comparison url.

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. Please note that if the perf results are neutral, you should likely undo the rollup=never given below by specifying rollup- to bors.

Importantly, though, if the results of this run are non-neutral do not roll this PR up -- it will mask other regressions or improvements in the roll up.

@bors rollup=never

compiler/rustc_mir_build/src/thir/pattern/_match.rs

varkor

This is a really nice series of improvements to the code; it feels a lot simpler now. Some of my comments may no longer make sense (or may apply to a different location), because I reviewed per-commit, but overall I have no issues with the refactoring itself: I just have a few comments about the comments and a couple of nits (I've been a little picky). After that, this should be good to merge. And that speed up is impressive ❤️

compiler/rustc_mir_build/src/thir/pattern/_match.rs

Nadrieril · 2020-10-28T19:21:45Z

Oh wow, no actual code comments, I must be getting good at this ^^. Thanks for the review!

@rustbot modify labels: -S-waiting-on-author +S-waiting-on-review

mark-i-m · 2020-10-28T19:51:58Z

This is an amazing PR. Refactoring, more lines removed than added, more comments, massive perf improvement. Great job!

varkor · 2020-10-28T21:50:30Z

Thanks, this looks great!

@bors r+ rollup=never

bors · 2020-10-28T21:50:32Z

📌 Commit 41a74ac has been approved by varkor

bors · 2020-10-29T01:37:55Z

⌛ Testing commit 41a74ac with merge f9187ad...

bors · 2020-10-29T03:57:47Z

☀️ Test successful - checks-actions
Approved by: varkor
Pushing f9187ad to master...

Before rust-lang#78430, string literals worked because `specialize_constructor` didn't actually care too much which constructor was passed to it unless needed. Since then, string literals are special cased and a bit hacky. I did not anticipate patterns for the `&str` type other than string literals, hence this bug. This makes string literals less hacky.

Fix rust-lang#78549 Before rust-lang#78430, this worked because `specialize_constructor` didn't actually care too much which constructor was passed to it unless needed. That PR however handles `&str` as a special case, and I did not anticipate patterns for the `&str` type other than string literals. I am not very confident there are not other similar oversights left, but hopefully only `&str` was different enough to break my assumptions. Fixes rust-lang#78549

Mark-Simulacrum · 2020-11-03T21:55:08Z

Excellent performance work, and followed by another PR further improving performance (#78553). Even better a refactoring of one of the most dense pieces of the compiler (at least IMO).

Nadrieril added 15 commits October 27, 2020 00:46

Split split_grouped_constructor into smaller functions

feb1e13

Use pat_constructor to simplify specialize_one_pattern

7c4f94b

Clarify specialization into two steps

6ad9f44

First is checking for constructor overlap, second is extracting the resulting fields.

Factor out the two specialization steps

c511955

Inline specialize_one_pattern

41e7ca4

Unify the two kinds of specialization by adding a Wildcard ctor

833089f

Cache head constructor in PatStack

1190e72

Since the constructor is recomputed a lot, caching is worth it.

Let MissingConstructors handle the subtleties of missing constructors

cdafd1e

Pass more things through PatCtxt

b49f907

This is even a perf improvement on the match-heavy benchmarks.

Recompute MissingConstructors when needed

c96bd28

This only happens in a slow (diagnostics) path, so the code clarity gain is worth it.

Unify the paths through is_useful

54fa702

Be honest about being able to list constructors

1fab669

The test change is because we used to treat `&str` like other `&T`s, ie as having a single constructor. That's not quite true though since we consider `&str` constants as atomic instead of refs to `str` constants.

Deduplicate work between splitting and subtraction

cd4c714

After splitting, subtraction becomes much simpler

Simplify slice splitting a bit

766ab78

rustbot added the A-exhaustiveness-checking Relating to exhaustiveness / usefulness checking of patterns label Oct 27, 2020

rust-highfive assigned varkor Oct 27, 2020

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 27, 2020

Nadrieril mentioned this pull request Oct 27, 2020

Rollup of 12 pull requests #78334

Merged

varkor reviewed Oct 27, 2020

View reviewed changes

varkor reviewed Oct 28, 2020

View reviewed changes

mark-i-m reviewed Oct 28, 2020

View reviewed changes

compiler/rustc_mir_build/src/thir/pattern/_match.rs Outdated Show resolved Hide resolved

varkor removed the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 28, 2020

Nadrieril force-pushed the taking-constructors-seriously2 branch from 9b42c7c to 41a74ac Compare October 28, 2020 19:08

rustbot 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 28, 2020

jyn514 added T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. C-cleanup Category: PRs that clean code up or issues documenting cleanup. I-compiletime Issue: Problems and improvements with respect to compile times. labels Oct 28, 2020

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 28, 2020

bors added the merged-by-bors This PR was explicitly merged by bors. label Oct 29, 2020

bors merged commit f9187ad into rust-lang:master Oct 29, 2020

rustbot added this to the 1.49.0 milestone Oct 29, 2020

This was referenced Oct 29, 2020

Fix unreachable sub-branch detection in or-patterns #78167

Merged

Rename overlapping_patterns lint #78242

Merged

Apply unused_doc_comments lint to inner items #78367

Closed

Nadrieril deleted the taking-constructors-seriously2 branch October 29, 2020 17:01

jyn514 mentioned this pull request Oct 30, 2020

ICE: trying to compare incompatible ctors while building rustfmt #78549

Closed

Nadrieril mentioned this pull request Oct 30, 2020

Fix exhaustiveness checking of strings #78553

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Clarify main code paths in exhaustiveness checking #78430

Clarify main code paths in exhaustiveness checking #78430

Nadrieril commented Oct 27, 2020 •

edited

Loading

jyn514 commented Oct 27, 2020

rust-timer commented Oct 27, 2020

bors commented Oct 27, 2020

bors commented Oct 27, 2020

rust-timer commented Oct 27, 2020

rust-timer commented Oct 27, 2020

varkor left a comment

Nadrieril commented Oct 28, 2020

mark-i-m commented Oct 28, 2020

varkor commented Oct 28, 2020

bors commented Oct 28, 2020

bors commented Oct 29, 2020

bors commented Oct 29, 2020

Mark-Simulacrum commented Nov 3, 2020

Clarify main code paths in exhaustiveness checking #78430

Clarify main code paths in exhaustiveness checking #78430

Conversation

Nadrieril commented Oct 27, 2020 • edited Loading

jyn514 commented Oct 27, 2020

rust-timer commented Oct 27, 2020

bors commented Oct 27, 2020

bors commented Oct 27, 2020

rust-timer commented Oct 27, 2020

rust-timer commented Oct 27, 2020

varkor left a comment

Choose a reason for hiding this comment

Nadrieril commented Oct 28, 2020

mark-i-m commented Oct 28, 2020

varkor commented Oct 28, 2020

bors commented Oct 28, 2020

bors commented Oct 29, 2020

bors commented Oct 29, 2020

Mark-Simulacrum commented Nov 3, 2020

Nadrieril commented Oct 27, 2020 •

edited

Loading