Replace `rustc_data_structures::thin_vec::ThinVec` with `thin_vec::ThinVec` by n...

Some changes occurred in src/librustdoc/clean/types.rs

cc @camelid

Awaiting bors try build completion.

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

Trying commit 2c4e3fd with merge ffb3c67...

Try build successful - checks-actions
Build commit: ffb3c67 (ffb3c6740056e998519f02c0dbb8072c2341c73f)

Finished benchmarking commit (ffb3c67): 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-review -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.

mean1	range	count2
Regressions (primary)	0.5%	[0.2%, 1.0%]	88
Regressions (secondary)	0.8%	[0.2%, 3.1%]	60
Improvements (primary)	-0.7%	[-1.0%, -0.5%]	2
Improvements (secondary)	-0.8%	[-1.0%, -0.4%]	16
All (primary)	0.5%	[-1.0%, 1.0%]	90

Max RSS (memory usage)

Results

Cycles

Results

Footnotes

1. the arithmetic mean of the percent change

2. number of relevant changes

@nnethercote I didn't check the code yet, just read your description but ... I was wondering if the perf results were the results you were expecting :). Can you comment on that?.

Yes, this is disappointing. I'll have to see if I can optimize thin-vec's code.

I already tried this once in rustdoc with similar results. One annoying thing with these global updates is that it's not easy to see the individual changes. It might very well be that e.g. 80 % of places where ThinVec is used were improved, but the rest regressed so bad that it swayed the total results.

I already tried this once in rustdoc with similar results.

Is there a PR?

(I already fixed a bug in thin-vec and made some small optimizations.)

Is there a PR?

Hmm, I found #92514, but that's about thin slices, not thin vecs. I probably didn't even send a PR and just tested it locally, but I remember using the thin_vec crate instead of the rustc ThinVec and not seeing any improvements. But that could have changed since then of course, and I think that I only tested rustdoc.

Hopefully Gankra/thin-vec#34 will fix most/all of the problems.

Awaiting bors try build completion.

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

Trying commit 745ad84 with merge 5c6507f...

Awaiting bors try build completion.

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

Trying commit 78f83f0 with merge ea65f30...

Try build successful - checks-actions
Build commit: ea65f30 (ea65f3049b7a96d203ed2fc38f2471bbb0e0cfa5)

Finished benchmarking commit (ea65f30): 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-review -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.

mean1	range	count2
Regressions (primary)	0.2%	[0.2%, 0.2%]	1
Regressions (secondary)	0.6%	[0.3%, 1.3%]	6
Improvements (primary)	-0.3%	[-0.3%, -0.3%]	1
Improvements (secondary)	-0.6%	[-0.8%, -0.4%]	11
All (primary)	-0.0%	[-0.3%, 0.2%]	2

Max RSS (memory usage)

Results

Cycles

Results

Footnotes

1. the arithmetic mean of the percent change

2. number of relevant changes

@spastorino The improvements now slightly outweigh the regressions. I don't entirely trust the last set of results, because I think there is some noise in these results. But I think it's close enough to performance-neutral to proceed.

@bors r+

Commit 78f83f0 has been approved by spastorino

It is now in the queue for this repository.

Testing commit 78f83f0 with merge eac6c33...

Test successful - checks-actions
Approved by: spastorino
Pushing eac6c33 to master...

Finished benchmarking commit (eac6c33): comparison URL.

Overall result: regressions - ACTION NEEDED

Next Steps: If you can justify the regressions found in this perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please open an issue or create a new PR that fixes the regressions, add a comment linking to the newly created issue or PR, and then add the perf-regression-triaged label to this PR.

@rustbot label: +perf-regression
cc @rust-lang/wg-compiler-performance

Instruction count

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

mean1	range	count2
Regressions (primary)	0.3%	[0.2%, 0.5%]	37
Regressions (secondary)	0.5%	[0.2%, 0.9%]	12
Improvements (primary)	-0.5%	[-0.6%, -0.4%]	4
Improvements (secondary)	-	-	0
All (primary)	0.2%	[-0.6%, 0.5%]	41

Max RSS (memory usage)

Results

Cycles

Results

Footnotes

1. the arithmetic mean of the percent change

2. number of relevant changes

Ugh, the post-merge perf run looked very different to the pre-merge perf run, despite there being no change to the code. Clearly a lot of noise here, and hard to know which results to trust. I'll see if I can get any follow-up improvements.

@nnethercote any ideas on follow ups here? The results seem like they're partially noise, but some would appear to be real issues.

Nothing right now, though I have a couple of experiments still to run. But thin_vec::ThinVec is a sufficiently useful tool when it comes to shrinking data structures in general that I'd like to keep it in place regardless. In general, the old rustc_data_structure one was good for cases where the vector is nearly almost always empty (e.g. attributes on expressions) but awful in any other case due to the double allocation. thin_vec has much wider applicability.