[Experiment] Rewrite exhaustiveness in one pass

[Nadrieril]

I've had this in mind for a looong time now. Arm reachability checking does a quadratic amount of work: for each arm it checks if it is reachable given the arms above it. This feels wasteful since we often end up re-exploring the same cases when we check for exhaustiveness.

This PR is an attempt to check reachability at the same time as exhaustiveness. This opens the door to a bunch of code simplifications I'm very excited.

Last time I tried I utterly destroyed performance. Until I get the right permissions to run perf in my dev environment, I'll run perf from here if that's ok.

r? @ghost

[Nadrieril]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 6e5c460dabf632f3cfc2b042be3a9ad6a43d7bc6 with merge f07aa9a3c6ed57d9d30e29332b5836d416ed417d...

[bors]

💔 Test failed - checks-actions

[Nadrieril]

I didn't know doc errors would stop bors 🙈

@bors try @rust-timer queue

[bors]

⌛ Trying commit 5735021ca97a51358b5d44d3df7d3ea936fa806c with merge 7b35e41a47c1bd7d4278379f89c90bd54038b471...

[bors]

☀️ Try build successful - checks-actions
Build commit: 7a6baddc69eb6fadc114345e5ca1eb81f2a07807 (7a6baddc69eb6fadc114345e5ca1eb81f2a07807)

[rust-timer]

Finished benchmarking commit (7a6baddc69eb6fadc114345e5ca1eb81f2a07807): comparison URL.

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. While you can manually mark this PR as fit for rollup, we strongly recommend not doing so since this PR may lead to changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please fix the regressions and do another perf run. If the next run shows neutral or positive results, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.6%	[0.4%, 0.9%]	3
Regressions ❌ (secondary)	3.4%	[0.3%, 4.7%]	8
Improvements ✅ (primary)	-0.8%	[-2.1%, -0.1%]	15
Improvements ✅ (secondary)	-0.3%	[-0.3%, -0.3%]	2
All ❌✅ (primary)	-0.6%	[-2.1%, 0.9%]	18

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	1.3%	[1.0%, 1.6%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.5%	[-3.5%, -3.5%]	1
All ❌✅ (primary)	-	-	0

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-7.5%	[-9.2%, -6.1%]	6
All ❌✅ (primary)	-	-	0

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 646.644s -> 647.317s (0.10%)

[Nadrieril]

This makes no sense I'm measuring a 14% improvement on match-stress locally :'(

[Kobzol]

Cycle wins are pretty nice: https://perf.rust-lang.org/compare.html?start=f3d597b31c0f101a02c230798afa31a36bdacbc6&end=7a6baddc69eb6fadc114345e5ca1eb81f2a07807&stat=cycles:u So I wouldn't mind the instruction regressions.

[nnethercote]

Walltime results are also nice, up to 8% wins on match-stress. Instruction counts are a decent metric in general, but the larger the change, the less reliable they become, and this seems like a fairly large change.

[Nadrieril]

Ohh nice I never checked the other metrics, ty both

[Nadrieril]

I lost momentum because of the perf issue, should come back to this this week or the next

[Kobzol]

@Nadrieril I'm curious, why did you close this? The perf. results on the match stress enum were really nice.

[Nadrieril]

I'm still excited about it and hope to finish it! But I lost momentum and life happened and I don't expect to be able to come back to it for weeks if not months :/. I'm closing it in the meantime for triage purposes
Thx for checking in btw, appreciate to know I'm not working unnoticed

[Kobzol]

I see. Take care!