Pre-compute MIR CFG caches for borrowck and other analyses

[cjgillot]

I was puzzled that #142390 introduces additional computations of CFG traversals: borrowck computes them, right?

It turns out that borrowck clones the MIR body, so doesn't share its cache with other analyses.

This PR:

• forces the computation of all caches in mir_promoted query;
• modifies region renumbering to avoid dropping that cache.

[rustbot]

r? @fee1-dead

rustbot has assigned @fee1-dead.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[rustbot]

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

[cjgillot]

@bors2 try @rust-timer queue

[rust-bors]

⌛ Trying commit e7e8f6d with merge 18218cc…

To cancel the try build, run the command @bors2 try cancel.

[rust-bors]

☀️ Try build successful (CI)
Build commit: 18218cc (18218cc5c664c3521f4769f46ef5e5e10128c92c, parent: 75e7cf5f85aad82331a38deff24845b63eaf30f3)

[rust-timer]

Finished benchmarking commit (18218cc): comparison URL.

Overall result: ❌ regressions - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	0.5%	[0.2%, 2.9%]	68
Regressions ❌ (secondary)	0.7%	[0.2%, 1.8%]	47
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.5%	[0.2%, 2.9%]	68

Max RSS (memory usage)

Results (primary 1.7%, secondary 1.3%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.7%	[0.9%, 2.8%]	12
Regressions ❌ (secondary)	2.5%	[2.0%, 3.0%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-1.2%	[-1.2%, -1.2%]	1
All ❌✅ (primary)	1.7%	[0.9%, 2.8%]	12

Cycles

Results (primary 1.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.1%	[0.7%, 1.4%]	2
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.1%	[0.7%, 1.4%]	2

Binary size

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

Bootstrap: 756.494s -> 756.086s (-0.05%)
Artifact size: 372.15 MiB -> 372.18 MiB (0.01%)

[cjgillot]

Perf deserves investigation. I really didn't expect mir_borrowck to get slower.

[cjgillot]

@bors2 try @rust-timer queue

[rust-bors]

⌛ Trying commit 3e77d40 with merge 98e2e6a…

To cancel the try build, run the command @bors2 try cancel.

[rust-bors]

☀️ Try build successful (CI)
Build commit: 98e2e6a (98e2e6aea9abc7e12f76e27d72aced95f45b5260, parent: 586ad391f5ee4519acc7cae340e34673bae762b1)

[rust-timer]

Finished benchmarking commit (98e2e6a): comparison URL.

Overall result: ❌ regressions - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	1.8%	[1.0%, 2.6%]	4
Regressions ❌ (secondary)	1.2%	[0.6%, 1.6%]	6
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.8%	[1.0%, 2.6%]	4

Max RSS (memory usage)

Results (primary 0.1%, secondary -2.5%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	3.1%	[0.5%, 6.5%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.4%	[-1.8%, -0.8%]	6
Improvements ✅ (secondary)	-2.5%	[-2.7%, -2.2%]	2
All ❌✅ (primary)	0.1%	[-1.8%, 6.5%]	9

Cycles

Results (secondary 1.2%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	1.2%	[1.2%, 1.2%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: 755.712s -> 754.846s (-0.11%)
Artifact size: 372.06 MiB -> 372.09 MiB (0.01%)

[cjgillot]

Rebased onto #142542
@bors2 try @rust-timer queue

[rust-bors]

⌛ Trying commit ba39fdf with merge 0e5f13e…

To cancel the try build, run the command @bors2 try cancel.

[rustbot]

This PR was rebased onto a different master commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[cjgillot]

@bors2 try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: 8e7d148 (8e7d14809b3504ba2f2a9e8caa6ae58e9c5fe86b, parent: ff6dc928c5e33ce8e65c6911a790b9efcb5ef53a)

[rust-timer]

Finished benchmarking commit (8e7d148): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	1.5%	[0.3%, 2.6%]	5
Regressions ❌ (secondary)	1.5%	[1.1%, 1.8%]	4
Improvements ✅ (primary)	-0.2%	[-0.4%, -0.1%]	49
Improvements ✅ (secondary)	-0.2%	[-0.3%, -0.0%]	35
All ❌✅ (primary)	-0.0%	[-0.4%, 2.6%]	54

Max RSS (memory usage)

Results (primary -2.9%, secondary -0.7%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	1.5%	[0.5%, 2.5%]	2
Improvements ✅ (primary)	-2.9%	[-6.8%, -0.8%]	3
Improvements ✅ (secondary)	-1.3%	[-2.1%, -0.6%]	8
All ❌✅ (primary)	-2.9%	[-6.8%, -0.8%]	3

Cycles

Results (secondary 0.2%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.3%	[2.1%, 2.4%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-4.0%	[-4.0%, -4.0%]	1
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: 471.991s -> 473.105s (0.24%)
Artifact size: 388.12 MiB -> 388.10 MiB (-0.01%)

[cjgillot]

I don't think I will do better here. This is an improvement on all but 2 benchmarks. I can't figure out why.
@rustbot ready

[fee1-dead]

I think this is fine. We're seeing enough wins for more common Rust code (serde, image, syn, etc) that I think it is fine to take the hit for the two outliers (tt-muncher and cranelift-codegen for some reason), I do think it could be because of cloning taking more instructions than what gets saved, but I don't think that is preventable.

@bors r+ rollup=never

[bors]

📌 Commit 42ff108 has been approved by fee1-dead

It is now in the queue for this repository.

[bors]

⌛ Testing commit 42ff108 with merge 28fad95...

[bors]

☀️ Test successful - checks-actions
Approved by: fee1-dead
Pushing 28fad95 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 f464759 (parent) -> 28fad95 (this PR)

Test differences

Show 2 test diffs

2 doctest diffs were found. These are ignored, as they are noisy.

Test dashboard

Run

cargo run --manifest-path src/ci/citool/Cargo.toml -- \
    test-dashboard 28fad9598975af7734be05116b3c5fc74b720d87 --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-x86_64-apple: 7239.0s -> 6005.0s (-17.0%)
2. test-various: 4099.6s -> 4729.1s (15.4%)
3. x86_64-gnu-gcc: 3053.2s -> 3465.6s (13.5%)
4. x86_64-gnu-llvm-20-1: 3227.1s -> 3622.4s (12.2%)
5. i686-gnu-nopt-1: 7316.6s -> 8196.5s (12.0%)
6. i686-gnu-2: 5624.2s -> 6245.1s (11.0%)
7. x86_64-gnu-llvm-20-3: 6452.3s -> 7156.3s (10.9%)
8. x86_64-gnu-llvm-20: 2540.1s -> 2805.8s (10.5%)
9. dist-aarch64-linux: 6579.6s -> 5964.6s (-9.3%)
10. dist-apple-various: 3666.2s -> 4004.2s (9.2%)

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 (28fad95): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Our benchmarks found a performance regression caused by this PR.
This might be an actual regression, but it can also be just noise.

Next Steps:

• If the regression was expected or you think it can be justified,
  please write a comment with sufficient written justification, and add
  @rustbot label: +perf-regression-triaged to it, to mark the regression as triaged.
• If you think that you know of a way to resolve the regression, try to create
  a new PR with a fix for the regression.
• If you do not understand the regression or you think that it is just noise,
  you can ask the @rust-lang/wg-compiler-performance working group for help (members of this group
  were already notified of this PR).

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	1.7%	[1.0%, 2.6%]	4
Regressions ❌ (secondary)	0.9%	[0.1%, 1.6%]	6
Improvements ✅ (primary)	-0.2%	[-0.3%, -0.1%]	15
Improvements ✅ (secondary)	-0.3%	[-0.4%, -0.2%]	13
All ❌✅ (primary)	0.2%	[-0.3%, 2.6%]	19

Max RSS (memory usage)

Results (primary -0.0%, secondary 0.5%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	3.0%	[3.0%, 3.0%]	1
Regressions ❌ (secondary)	2.9%	[1.6%, 4.1%]	2
Improvements ✅ (primary)	-1.6%	[-2.1%, -1.0%]	2
Improvements ✅ (secondary)	-1.1%	[-1.9%, -0.6%]	3
All ❌✅ (primary)	-0.0%	[-2.1%, 3.0%]	3

Cycles

Results (primary -2.0%, secondary -3.3%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.0%	[-2.0%, -2.0%]	1
Improvements ✅ (secondary)	-3.3%	[-3.3%, -3.3%]	1
All ❌✅ (primary)	-2.0%	[-2.0%, -2.0%]	1

Binary size

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

Bootstrap: 475.372s -> 475.32s (-0.01%)
Artifact size: 390.39 MiB -> 390.35 MiB (-0.01%)

[lqd]

cranelift-codegen is one of the benchmarks where borrowck takes the longest, is it expected that this change impacts it so negatively?

[Mark-Simulacrum]

I think this is fine. We're seeing enough wins for more common Rust code
(serde, image, syn, etc) that I think it is fine to take the hit for the two
outliers (tt-muncher and cranelift-codegen for some reason), I do think it
could be because of cloning taking more instructions than what gets saved,
but I don't think that is preventable.

#142540 (comment)

I think I'm broadly in agreement with this summary of the results. Definitely would be interested in figuring out why we have the outliers we do, though, and seems plausible there's an opportunity waiting there...