unexpand: save stack size - perf +6.98%

[oech3]

Replace a stack array with vec to avoid stack spill seen at rust-lang/rust#148670 .

Probally fixes #10654

[codspeed-hq]

Merging this PR will not alter performance

✅ 2 untouched benchmarks
⏩ 354 skipped benchmarks¹

---

_{Comparing oech3:unexpand-stack (266ef23) with main (b2619df)²}

Footnotes

1. 354 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on main (717140a) during the generation of this report, so b2619df was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/misc/io-errors. tests/misc/io-errors is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/tty/tty-eof (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/pr/bounded-memory (passes in this run but fails in the 'main' branch)
Note: The gnu test tests/seq/seq-epipe is now being skipped but was previously passing.
Congrats! The gnu test tests/tail/tail-n0f is now passing!

[sylvestre]

i wasn't expecting such wins for this kind of changes

[xtqqczze]

I don't understand why changing stack allocation to heap allocation is a performance improvement. Additionally, it may be possible that this isn't an improvement on targets other than x86_64-unknown-linux-gnu.

[oech3]

I'm OK to have cfg block.

[xtqqczze]

It would be good to understand why there is a performance difference; if necessary this could be reported upstream.

[oech3]

done at rust-lang/rust#148670

[xtqqczze]

done at rust-lang/rust#148670

Please add a comment linking to this issue; it would also have been helpful to include it in the PR description.

[xtqqczze]

I guess it would be good to explain in the comment that the regression is specifically with codegen-units=1 + LTO.

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/timeout/timeout-group. tests/timeout/timeout-group is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/tty/tty-eof (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/expand/bounded-memory is now passing!

[xtqqczze]

I'm OK to have cfg block.

Like efd0f0c, or just x86_64-unknown-linux-gnu?

[oech3]

I think cfg is overkill considering Linux-x64 is most common target (until targets significally decreasing perf was found).

[xtqqczze]

This change is a workaround for an upstream issue and should not be applied to targets where the issue has not been demonstrated.

[oech3]

I guess this is no longer needed as #11806 fixed stack spill. But I'll check CodSpeed few times before closing.

[xtqqczze]

@oech3 Thanks for looking into this!