Defer extra liveness calculation for Polonius Alpha

[jackh726]

View all comments

Best reviewed by commit.

This moves liveness calculation of NLL-boring/Polonius-relevant locals to be lazy. This allows us to skip unnecessary work.

r? lqd

[jackh726]

@bors try @rust-time queue

[jackh726]

@rust-timer queue

[jackh726]

Disclosure: An LLM was used for experiments here, but all code here is written fresh.

[rust-bors]

💔 Test for 254d008 failed: CI. Failed job:

• try - dist-x86_64-linux-quick (web logs, enhanced plaintext logs)

[jackh726]

@bors try

[rust-bors]

☀️ Try build successful (CI)
Build commit: f904a04 (f904a046d815a8f17021ba39511de661f74c1612)
Base parent: 344f790 (344f7902949345394fa40a5d7dda31f012ccbc0d)

[rust-timer]

Finished benchmarking commit (f904a04): comparison URL.

Overall result: ❌✅ regressions and improvements - please read:

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

Next, please: If you can, justify the regressions found in this try perf run in writing along with @rustbot label: +perf-regression-triaged. If not, fix the regressions and do another perf run. Neutral or positive results will clear the label automatically.

@bors rollup=never rustc-perf
@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)	0.4%	[0.3%, 0.5%]	11
Regressions ❌ (secondary)	1.2%	[0.1%, 3.5%]	20
Improvements ✅ (primary)	-3.3%	[-11.9%, -0.2%]	18
Improvements ✅ (secondary)	-4.5%	[-11.6%, -0.1%]	19
All ❌✅ (primary)	-1.9%	[-11.9%, 0.5%]	29

Max RSS (memory usage)

Results (primary -2.0%, 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)	2.9%	[0.8%, 9.2%]	16
Improvements ✅ (primary)	-2.0%	[-2.0%, -2.0%]	1
Improvements ✅ (secondary)	-4.4%	[-5.3%, -3.1%]	5
All ❌✅ (primary)	-2.0%	[-2.0%, -2.0%]	1

Cycles

Results (primary -3.6%, secondary -5.5%)

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)	4.6%	[4.2%, 5.0%]	2
Improvements ✅ (primary)	-3.6%	[-4.4%, -3.0%]	6
Improvements ✅ (secondary)	-6.9%	[-19.1%, -2.3%]	14
All ❌✅ (primary)	-3.6%	[-4.4%, -3.0%]	6

Binary size

Results (secondary -0.0%)

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)	-	-	0
Improvements ✅ (secondary)	-0.0%	[-0.0%, -0.0%]	4
All ❌✅ (primary)	-	-	0

Bootstrap: 475.673s -> 475.822s (0.03%)
Artifact size: 403.60 MiB -> 402.91 MiB (-0.17%)

[jackh726]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: 5c7880f (5c7880fc42042067782ea34311ccc9d8b517623a)
Base parent: feaadee (feaadeeaca7db0594da854e7c8c07495341c7439)

[rust-timer]

Finished benchmarking commit (5c7880f): comparison URL.

Overall result: ❌✅ regressions and improvements - please read:

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

Next, please: If you can, justify the regressions found in this try perf run in writing along with @rustbot label: +perf-regression-triaged. If not, fix the regressions and do another perf run. Neutral or positive results will clear the label automatically.

@bors rollup=never rustc-perf
@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)	0.2%	[0.2%, 0.2%]	1
Regressions ❌ (secondary)	2.5%	[2.3%, 2.7%]	6
Improvements ✅ (primary)	-1.6%	[-4.5%, -0.1%]	31
Improvements ✅ (secondary)	-0.8%	[-4.2%, -0.1%]	19
All ❌✅ (primary)	-1.5%	[-4.5%, 0.2%]	32

Max RSS (memory usage)

Results (primary 1.8%)

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

	mean	range	count
Regressions ❌ (primary)	1.8%	[1.8%, 1.8%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.8%	[1.8%, 1.8%]	1

Cycles

Results (primary -2.7%, secondary -1.5%)

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)	2.9%	[2.9%, 2.9%]	1
Improvements ✅ (primary)	-2.7%	[-2.7%, -2.7%]	1
Improvements ✅ (secondary)	-3.7%	[-4.6%, -2.9%]	2
All ❌✅ (primary)	-2.7%	[-2.7%, -2.7%]	1

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 502.551s -> 497.776s (-0.95%)
Artifact size: 406.86 MiB -> 408.96 MiB (0.52%)

[jackh726]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: a51a48c (a51a48c6f77b69d10ec98590efd55f70e2d90fd9)
Base parent: d287eb7 (d287eb7a292caa8abbe12051dce2ee707fe03200)

[rust-timer]

Finished benchmarking commit (a51a48c): comparison URL.

Overall result: ❌✅ regressions and improvements - please read:

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

Next, please: If you can, justify the regressions found in this try perf run in writing along with @rustbot label: +perf-regression-triaged. If not, fix the regressions and do another perf run. Neutral or positive results will clear the label automatically.

@bors rollup=never rustc-perf
@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
Regressions ❌ (secondary)	1.9%	[0.2%, 2.8%]	9
Improvements ✅ (primary)	-1.5%	[-4.5%, -0.1%]	33
Improvements ✅ (secondary)	-1.2%	[-4.2%, -0.2%]	10
All ❌✅ (primary)	-1.5%	[-4.5%, -0.1%]	33

Max RSS (memory usage)

Results (primary -0.4%)

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

	mean	range	count
Regressions ❌ (primary)	4.9%	[4.9%, 4.9%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.1%	[-3.5%, -2.7%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.4%	[-3.5%, 4.9%]	3

Cycles

Results (primary -2.1%, secondary 3.4%)

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)	3.4%	[3.3%, 3.5%]	3
Improvements ✅ (primary)	-2.1%	[-2.1%, -2.1%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-2.1%	[-2.1%, -2.1%]	1

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 503.997s -> 498.581s (-1.07%)
Artifact size: 408.93 MiB -> 409.02 MiB (0.02%)

[lqd]

We were wondering about the perf results, since we split a handful of commits from this PR and landed them already:

• in the first perf run in this serde-check-full went from 6.3 GInstr to 5.6 GInstr
• in Cache block seeking results for initialized_at_terminator and _exit #162450, it went from 6 .1 to 5.8
• in the most recent run in this PR, it went from 5.8 to 5.5
• serde-nll on main is currently at 5.48

In the last run there are also more primary improvements, less secondary improvements, less regressions (only 1, on the wg-grammar stress test, with a similar range on all runs), compared to the previous runs.

Overall this is looking good

• ripgrep-nll on main is at 1.41, polonius here is at 1.45
• clap-derive-nll on main is at 2.11, polonius here is at 2.27

And after this PR, we'll also have the next big change about computing multiple loans at a time rather per loan, so we're still on a good path to greatly reduce the overhead.