Modify the buffer position directly when reading leb128 values.

[nnethercote]

It's a small but clear performance win.

r? @michaelwoerister

[nnethercote]

@bors try @rust-timer queue

[rust-timer]

Awaiting bors try build completion.

@rustbot label: +S-waiting-on-perf

[bors]

⌛ Trying commit 5f549d9 with merge 50c7ddde3ce168b8c3ba9a790e49d8df73d08fdd...

[bors]

☀️ Try build successful - checks-actions
Build commit: 50c7ddde3ce168b8c3ba9a790e49d8df73d08fdd (50c7ddde3ce168b8c3ba9a790e49d8df73d08fdd)

[rust-timer]

Queued 50c7ddde3ce168b8c3ba9a790e49d8df73d08fdd with parent cfa4ac6, future comparison URL.

[rust-timer]

Finished benchmarking commit (50c7ddde3ce168b8c3ba9a790e49d8df73d08fdd): comparison url.

Summary: This change led to small relevant improvements 🎉 in compiler performance.

• Small improvement in instruction counts (up to -1.0% on incr-unchanged builds of ucd)

If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf.

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 led to changes in compiler perf.

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

[nnethercote]

The CI results show universal instruction count improvements, but they are smaller than what I saw locally, and spread differently across the benchmarks. Here are the top 10 check results on CI:

Benchmark & Profile	Scenario	% Change	Significance Factor
ucd check	incr-unchanged	-1.03%?	2.46x
tuple-stress check	incr-unchanged	-1.00%	14.85x
coercions check	incr-unchanged	-0.64%	6.49x
coercions check	incr-patched: println	-0.58%	5.73x
html5ever check	incr-patched: println	-0.52%?	2.08x
tokio-webpush-simple check	full	-0.35%	1.99x
webrender-wrench check	incr-unchanged	-0.34%	1.74x
wg-grammar check	incr-unchanged	-0.33%	1.93x
tokio-webpush-simple check	incr-full	-0.27%	1.69x
regex check	incr-patched: Job	-0.27%	1.51x

And here they from my local measurements (where I only did check builds):

Benchmark & Profile	Scenario	% Change	Significance Factor
helloworld check	full	-1.91%	9.56x
issue-46449 check	incr-unchanged	-1.86%	9.30x
issue-46449 check	incr-patched: u8 3072	-1.74%	8.68x
helloworld check	incr-full	-1.72%	8.62x
helloworld check	incr-unchanged	-1.69%	8.43x
deeply-nested check	incr-unchanged	-1.60%	8.02x
deeply-nested-closures check	incr-unchanged	-1.57%	7.87x
await-call-tree check	incr-unchanged	-1.55%	7.74x
helloworld check	incr-patched: println	-1.54%	7.72x
issue-46449 check	full	-1.53%	7.63x

I feel like the match-ups between my local measurements and CI measurements are much weaker than they used to be. I wonder if that's because I don't do PGO locally.

[michaelwoerister]

Thanks, @nnethercote! Looks good to me.

@bors r+

I wonder if that's because I don't do PGO locally.

Yeah, I suspect that that plays into it.

[bors]

📌 Commit 5f549d9 has been approved by michaelwoerister

[michaelwoerister]

I just saw that #92604 already includes the changes in this PR. Let's take that instead.

@bors r-

[michaelwoerister]

Closing in favor of #92604