Avoid checking the edition as much as possible

[vincenzopalazzo]

Inside #123865, we are adding support for the new semantics for expr2024, but we have noted a performance issue.

While talking with @eholk, we realized there is a redundant check for each token regarding an edition. This commit moves the edition check to the end, avoiding some extra checks that can slow down compilation time.

However, we should keep this issue under observation because we may want to improve the edition check if we are unable to significantly improve compiler performance.

r? ghost

[vincenzopalazzo]

@bors try @rust-timer queue

You can follow the discussion on zulip

[bors]

@vincenzopalazzo: 🔑 Insufficient privileges: not in try users

[eholk]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 252f83a with merge d693eba...

[matthiaskrgr]

🤔 I would assume that PGO already does such kind of optimization on its own.

[bors]

☀️ Try build successful - checks-actions
Build commit: d693eba (d693ebad7a52d8ab0e267c5e72e5983aa878757a)

[rust-timer]

Finished benchmarking commit (d693eba): comparison URL.

Overall result: ✅ improvements - no 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.

@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
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.3%	[-0.4%, -0.2%]	8
Improvements ✅ (secondary)	-0.3%	[-0.4%, -0.3%]	4
All ❌✅ (primary)	-0.3%	[-0.4%, -0.2%]	8

Max RSS (memory usage)

Results (primary 3.9%)

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)	3.9%	[3.9%, 3.9%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	3.9%	[3.9%, 3.9%]	1

Cycles

Results (primary -1.1%, secondary 2.0%)

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)	2.7%	[2.0%, 4.0%]	9
Improvements ✅ (primary)	-1.1%	[-1.4%, -0.9%]	11
Improvements ✅ (secondary)	-1.2%	[-1.5%, -0.9%]	2
All ❌✅ (primary)	-1.1%	[-1.4%, -0.9%]	11

Binary size

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

Bootstrap: 672.771s -> 666.635s (-0.91%)
Artifact size: 318.83 MiB -> 318.84 MiB (0.00%)

[eholk]

Comparing these results with the results from the last PR, it looks like this more than makes up for the regressions in html5ever, but only gets back some of the performance for ttmuncher.

[workingjubilee]

🤔 I would assume that PGO already does such kind of optimization on its own.

It is incorrect to reorder a || b into b || a, or the same with &&, in the general case, whereas a | b can be ordered in any direction.

[workingjubilee]

And yes, the optimization may be permissible in this case, but LLVM got slightly more skittish about reordering those after it got burned by doing the optimization unsoundly.

[fmease]

CC the results from the full revert (#125389)

[vincenzopalazzo]

Sorry for the delay!

Well, we've already made some improvements with @eholk's suggestion, right? Why don't we try to merge this PR and track the performance issue with a tracking issue?

I think we already have several PRs for expr2024, so maybe it's worth looking at this later to see if we need some other kind of optimization that allows us cheap access to the edition?

r? fmease

[fmease]

@bors r+

[bors]

📌 Commit 36d5fc9 has been approved by fmease

It is now in the queue for this repository.

[bors]

⌛ Testing commit 36d5fc9 with merge a6416d8...

[bors]

☀️ Test successful - checks-actions
Approved by: fmease
Pushing a6416d8 to master...

[rust-timer]

Finished benchmarking commit (a6416d8): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -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
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.4%	[-0.4%, -0.3%]	6
Improvements ✅ (secondary)	-0.4%	[-0.5%, -0.3%]	4
All ❌✅ (primary)	-0.4%	[-0.4%, -0.3%]	6

Max RSS (memory usage)

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

Cycles

Results (primary 2.2%, secondary 2.5%)

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)	2.2%	[1.8%, 2.4%]	3
Regressions ❌ (secondary)	3.1%	[2.0%, 4.5%]	10
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.6%	[-2.6%, -2.6%]	1
All ❌✅ (primary)	2.2%	[1.8%, 2.4%]	3

Binary size

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

Bootstrap: 669.121s -> 667.68s (-0.22%)
Artifact size: 318.92 MiB -> 318.93 MiB (0.00%)