cstore: Use IndexSet as backing store for postorder dependencies

[osiewicz]

<rustc_metadata::creader::CStore>::push_dependencies_in_postorder showed up in new benchmarks from rust-lang/rustc-perf#2143, hence I gave it a shot to remove an obvious O(n) there.

r? nnethercote

[bjorn3]

@bors try @rust-timer queue

[bors]

⌛ Testing commit 9b3be25 with merge 0f0df11...

[Kobzol]

Bors got confused as the try build was started just a as a commit was pushed here. I'll cancel the workflow.

[bors]

💔 Test failed - checks-actions

[Kobzol]

@bors try

[bors]

⌛ Trying commit 9b3be25 with merge a8791be...

[bors]

☀️ Try build successful - checks-actions
Build commit: a8791be (a8791bea639ea3cb555d4c410091517cd0e406f3)

[rust-timer]

Finished benchmarking commit (a8791be): 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 the most reliable metric that we have; it was used to determine the overall result at the top of this comment. However, even this metric can sometimes exhibit noise.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.2%	[-5.2%, -0.4%]	12
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results (primary -1.9%, secondary -2.4%)

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)	0.6%	[0.6%, 0.7%]	3
Improvements ✅ (primary)	-1.9%	[-3.4%, -1.1%]	7
Improvements ✅ (secondary)	-2.6%	[-5.1%, -0.5%]	56
All ❌✅ (primary)	-1.9%	[-3.4%, -1.1%]	7

Cycles

Results (secondary -1.0%)

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)	0.6%	[0.5%, 0.7%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-1.3%	[-2.4%, -0.4%]	11
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: 777.423s -> 777.791s (0.05%)
Artifact size: 368.47 MiB -> 368.48 MiB (0.00%)

[bjorn3]

Significant improvements for large-workspace as expected and almost no effect on all other benchmarks. Interestingly this also improves max-rss a bit across the board.

[nnethercote]

Interestingly this also improves max-rss a bit across the board.

I don't understand that, and I'm very suspicious. An IndexSet can only be larger than an equivalent Vec. Let's see if the results hold up on the post-merge perf run.

Anyway, the icount results look great. Nice work!

@bors r+

[bors]

📌 Commit 5e61f5e has been approved by nnethercote

It is now in the queue for this repository.

[osiewicz]

This PR also changed the call in query system from alloc_slice to alloc_from_iter - could that be the source of a RSS difference?

[nnethercote]

Seems unlikely. I did a local before/after profile with DHAT on helloworld (one of the most-affected benchmarks) and didn't see any difference between the profiles.

[bors]

⌛ Testing commit 5e61f5e with merge ba55b7c...

[bors]

☀️ Test successful - checks-actions
Approved by: nnethercote
Pushing ba55b7c to master...

[github-actions]

What is this?

This is an experimental post-merge analysis report that shows differences in test outcomes between the merged PR and its parent PR.

Comparing 13a4540 (parent) -> ba55b7c (this PR)

Test differences

No test diffs found

Test dashboard

Run

cargo run --manifest-path src/ci/citool/Cargo.toml -- \
    test-dashboard ba55b7ce3cd931af7123e334647584f6fbdb3c20 --output-dir test-dashboard

And then open test-dashboard/index.html in your browser to see an overview of all executed tests.

Job duration changes

1. dist-aarch64-linux: 7816.5s -> 5466.3s (-30.1%)
2. x86_64-apple-2: 5688.5s -> 4429.5s (-22.1%)
3. dist-apple-various: 8386.8s -> 6902.8s (-17.7%)
4. dist-aarch64-apple: 6027.8s -> 5427.5s (-10.0%)
5. aarch64-apple: 5203.2s -> 4888.6s (-6.0%)
6. x86_64-apple-1: 6776.7s -> 7181.9s (6.0%)
7. dist-x86_64-musl: 7310.6s -> 7718.5s (5.6%)
8. dist-arm-linux-musl: 6008.2s -> 5732.2s (-4.6%)
9. dist-x86_64-apple: 9122.9s -> 9525.3s (4.4%)
10. dist-i686-msvc: 7234.6s -> 6916.8s (-4.4%)

How to interpret the job duration changes?

Job durations can vary a lot, based on the actual runner instance
that executed the job, system noise, invalidated caches, etc. The table above is provided
mostly for t-infra members, for simpler debugging of potential CI slow-downs.

[rust-timer]

Finished benchmarking commit (ba55b7c): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -perf-regression

Instruction count

This is the most reliable metric that we have; it was used to determine the overall result at the top of this comment. However, even this metric can sometimes exhibit noise.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.3%	[-5.2%, -0.4%]	11
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results (secondary -0.1%)

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)	0.9%	[0.5%, 1.8%]	6
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.0%	[-3.0%, -0.9%]	3
All ❌✅ (primary)	-	-	0

Cycles

Results (primary -1.6%, secondary -1.4%)

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)	1.0%	[0.5%, 2.2%]	5
Improvements ✅ (primary)	-1.6%	[-1.9%, -1.4%]	3
Improvements ✅ (secondary)	-2.0%	[-4.4%, -0.4%]	19
All ❌✅ (primary)	-1.6%	[-1.9%, -1.4%]	3

Binary size

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

Bootstrap: 775.153s -> 774.617s (-0.07%)
Artifact size: 372.28 MiB -> 372.30 MiB (0.01%)