Make skip_whitespace do a single pass (with bytes)

[hkBst]

Probably better alternative to #137027.

r? @nnethercote

[jieyouxu]

@bors try @rust-timer queue

[bors]

⌛ Trying commit d4feb08 with merge f62fb4e...

[bors]

☀️ Try build successful - checks-actions
Build commit: f62fb4e (f62fb4e8b41a377accf89c9d5ea21df1c1333251)

[rust-timer]

Finished benchmarking commit (f62fb4e): comparison URL.

Overall result: no relevant changes - 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 benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results (primary -1.6%)

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)	-1.6%	[-1.6%, -1.6%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.6%	[-1.6%, -1.6%]	1

Cycles

Results (primary -4.0%)

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)	-	-	0
Improvements ✅ (primary)	-4.0%	[-5.9%, -1.6%]	6
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-4.0%	[-5.9%, -1.6%]	6

Binary size

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

Bootstrap: 773.063s -> 774.969s (0.25%)
Artifact size: 362.34 MiB -> 362.33 MiB (-0.00%)

[nnethercote]

@hkBst: is this ready for review?

[hkBst]

@nnethercote I've now added @folkertdev's changes which look like an improvement, so another perf run would be good, but yes in principle this is ready for review.

[nnethercote]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 258115e with merge 665ef7c...

[bors]

☀️ Try build successful - checks-actions
Build commit: 665ef7c (665ef7c7cc094b58b3753cbd33b8ce51dfaf4142)

[rust-timer]

Finished benchmarking commit (665ef7c): comparison URL.

Overall result: no relevant changes - 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 benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

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

Cycles

Results (secondary 1.3%)

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)	5.7%	[5.7%, 5.7%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.2%	[-3.2%, -3.2%]	1
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: 772.938s -> 773.607s (0.09%)
Artifact size: 360.27 MiB -> 360.30 MiB (0.01%)

[folkertdev]

As a general question, would we ever expect (small) parser changes to show up as significant in the overall performance benchmarks? Surely that gets drowned out by semantic analysis and (if used) codegen?

[hkBst]

As a general question, would we ever expect (small) parser changes to show up as significant in the overall performance benchmarks? Surely that gets drowned out by semantic analysis and (if used) codegen?

This is probably too specific, as it only affects strings that use whitespace continuation, and makes that part a bit faster (probably?). But small changes can have like a -0.1% effect.

[nnethercote]

As a general question, would we ever expect (small) parser changes to show up as significant in the overall performance benchmarks? Surely that gets drowned out by semantic analysis and (if used) codegen?

They can definitely show up in incremental builds that have few or no changes, because parsing isn't incremental. And also check builds. See #136931 for a recent related example that had to be abandoned for perf regressions.

[hkBst]

I've been running some performance tests, and even in the best case scenario, this change seems to perform 25% worse than the current double traversal... so I may have to pivot to documenting that...