Sandwich MIR optimizations between DSE. #119672

cjgillot · 2024-01-06T19:39:57Z

This PR reorders MIR optimization passes in an attempt to increase their efficiency.

Stop running CopyProp before GVN, it's useless as GVN will do the same thing anyway. Instead, we perform CopyProp at the end of the pipeline, to ensure we do not emit copy/move chains.
Run DSE before GVN, as it increases the probability to have single-assignment locals.
Run DSE after the final CopyProp to turn copies into moves.

cjgillot · 2024-01-06T19:40:07Z

bors · 2024-01-06T19:41:17Z

⌛ Trying commit bee69ed with merge da2b34d...

Sandwich MIR optimizations between DSE. This PR reorders MIR optimization passes in an attempt to increase their efficiency. - Stop running CopyProp before GVN, it's useless as GVN will do the same thing anyway. Instead, we perform CopyProp at the end of the pipeline, to ensure we do not emit copy/move chains. - Run DSE before GVN, as it increases the probability to have single-assignment locals. - Run DSE after the final CopyProp to turn copies into moves. r? `@ghost`

bors · 2024-01-06T21:07:31Z

☀️ Try build successful - checks-actions
Build commit: da2b34d (da2b34d645102a105976e2818671f2be9fc7cf06)

rust-timer · 2024-01-06T22:27:58Z

Finished benchmarking commit (da2b34d): 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-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.

	mean	range	count
Regressions ❌ (primary)	0.8%	[0.2%, 4.7%]	19
Regressions ❌ (secondary)	0.4%	[0.2%, 0.8%]	12
Improvements ✅ (primary)	-1.0%	[-2.2%, -0.2%]	28
Improvements ✅ (secondary)	-0.8%	[-1.9%, -0.2%]	13
All ❌✅ (primary)	-0.2%	[-2.2%, 4.7%]	47

Max RSS (memory usage)

Results

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)	4.8%	[1.9%, 14.4%]	5
Regressions ❌ (secondary)	3.8%	[1.7%, 9.5%]	4
Improvements ✅ (primary)	-1.6%	[-2.4%, -0.1%]	4
Improvements ✅ (secondary)	-5.1%	[-7.8%, -3.7%]	3
All ❌✅ (primary)	2.0%	[-2.4%, 14.4%]	9

Cycles

Results

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)	2.8%	[1.1%, 4.5%]	2
Regressions ❌ (secondary)	3.0%	[2.2%, 3.9%]	2
Improvements ✅ (primary)	-1.9%	[-2.4%, -1.4%]	8
Improvements ✅ (secondary)	-2.2%	[-2.3%, -2.0%]	3
All ❌✅ (primary)	-1.0%	[-2.4%, 4.5%]	10

Binary size

Results

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)	0.7%	[0.0%, 2.2%]	14
Regressions ❌ (secondary)	0.0%	[0.0%, 0.0%]	1
Improvements ✅ (primary)	-0.3%	[-1.0%, -0.0%]	69
Improvements ✅ (secondary)	-0.5%	[-1.1%, -0.2%]	21
All ❌✅ (primary)	-0.1%	[-1.0%, 2.2%]	83

Bootstrap: 668.367s -> 667.451s (-0.14%)
Artifact size: 308.49 MiB -> 308.66 MiB (0.05%)

bors · 2024-01-09T17:51:23Z

☔ The latest upstream changes (presumably #119767) made this pull request unmergeable. Please resolve the merge conflicts.

bors · 2024-01-11T07:07:33Z

☔ The latest upstream changes (presumably #119677) made this pull request unmergeable. Please resolve the merge conflicts.

The full pass is run short after.

cjgillot · 2024-01-12T21:55:18Z

Now that #119677 is merged.
@bors try @rust-timer queue

bors · 2024-01-12T21:56:31Z

⌛ Trying commit bc35ee4 with merge cc8825e...

Sandwich MIR optimizations between DSE. This PR reorders MIR optimization passes in an attempt to increase their efficiency. - Stop running CopyProp before GVN, it's useless as GVN will do the same thing anyway. Instead, we perform CopyProp at the end of the pipeline, to ensure we do not emit copy/move chains. - Run DSE before GVN, as it increases the probability to have single-assignment locals. - Run DSE after the final CopyProp to turn copies into moves. r? `@ghost`

bors · 2024-01-12T23:21:29Z

☀️ Try build successful - checks-actions
Build commit: cc8825e (cc8825e3c98ad94907c9ea8483edd15220baa7e3)

rust-timer · 2024-01-13T00:41:31Z

Finished benchmarking commit (cc8825e): 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-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.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.1%, 1.2%]	12
Regressions ❌ (secondary)	0.5%	[0.2%, 1.8%]	10
Improvements ✅ (primary)	-1.1%	[-2.1%, -0.2%]	27
Improvements ✅ (secondary)	-1.0%	[-2.2%, -0.2%]	9
All ❌✅ (primary)	-0.5%	[-2.1%, 1.2%]	39

Max RSS (memory usage)

Results

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)	2.4%	[0.9%, 3.6%]	6
Regressions ❌ (secondary)	2.6%	[2.1%, 3.1%]	2
Improvements ✅ (primary)	-4.0%	[-18.0%, -0.4%]	9
Improvements ✅ (secondary)	-8.4%	[-8.4%, -8.4%]	1
All ❌✅ (primary)	-1.4%	[-18.0%, 3.6%]	15

Cycles

Results

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)	1.9%	[1.9%, 1.9%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.6%	[-2.6%, -0.5%]	9
Improvements ✅ (secondary)	-1.9%	[-3.3%, -0.9%]	5
All ❌✅ (primary)	-1.2%	[-2.6%, 1.9%]	10

Binary size

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

Bootstrap: 666.309s -> 663.818s (-0.37%)
Artifact size: 308.10 MiB -> 308.22 MiB (0.04%)

rustbot · 2024-01-13T00:47:38Z

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

cjgillot · 2024-01-13T00:51:30Z

Perf results: we get some nice speedups, near 2% for cargo, clap & ripgrep, which confirm the earlier perf run. On average, small improvements and small regressions balance out.
I propose to go ahead and merge this.
r? wg-mir-opt

oli-obk · 2024-01-15T13:33:32Z

@bors r+

bors · 2024-01-15T13:33:35Z

📌 Commit bc35ee4 has been approved by oli-obk

It is now in the queue for this repository.

bors · 2024-01-16T00:04:20Z

⌛ Testing commit bc35ee4 with merge 00b705e...

Sandwich MIR optimizations between DSE. This PR reorders MIR optimization passes in an attempt to increase their efficiency. - Stop running CopyProp before GVN, it's useless as GVN will do the same thing anyway. Instead, we perform CopyProp at the end of the pipeline, to ensure we do not emit copy/move chains. - Run DSE before GVN, as it increases the probability to have single-assignment locals. - Run DSE after the final CopyProp to turn copies into moves. r? `@ghost`

bors · 2024-01-16T04:19:21Z

💥 Test timed out

rust-log-analyzer · 2024-01-16T05:18:32Z

A job failed! Check out the build log: (web) (plain)

Click to see the possible cause of the failure (guessed by this bot)

oli-obk · 2024-01-16T08:30:11Z

@bors retry apple builder timeout

bors · 2024-01-16T11:34:19Z

⌛ Testing commit bc35ee4 with merge fa0dc20...

bors · 2024-01-16T13:31:51Z

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing fa0dc20 to master...

rust-timer · 2024-01-16T15:00:23Z

Finished benchmarking commit (fa0dc20): comparison URL.

Overall result: ❌✅ regressions and improvements - 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.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.2%, 1.4%]	14
Regressions ❌ (secondary)	0.5%	[0.2%, 2.7%]	14
Improvements ✅ (primary)	-1.0%	[-2.2%, -0.2%]	31
Improvements ✅ (secondary)	-0.9%	[-2.2%, -0.2%]	10
All ❌✅ (primary)	-0.4%	[-2.2%, 1.4%]	45

Max RSS (memory usage)

Results

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)	2.4%	[2.4%, 2.4%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-4.0%	[-7.9%, -0.7%]	6
Improvements ✅ (secondary)	-5.2%	[-8.2%, -3.1%]	3
All ❌✅ (primary)	-3.1%	[-7.9%, 2.4%]	7

Cycles

Results

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)	1.4%	[1.4%, 1.4%]	2
Regressions ❌ (secondary)	2.0%	[1.8%, 2.3%]	4
Improvements ✅ (primary)	-1.8%	[-2.7%, -1.1%]	11
Improvements ✅ (secondary)	-2.5%	[-3.1%, -2.0%]	3
All ❌✅ (primary)	-1.3%	[-2.7%, 1.4%]	13

Binary size

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

Bootstrap: 666.062s -> 664.881s (-0.18%)
Artifact size: 308.24 MiB -> 308.33 MiB (0.03%)

lqd · 2024-01-16T15:05:22Z

As seen in the previous runs: some nice wins on bigger benchmarks, and overall gains outweigh the few losses.

@rustbot label: +perf-regression-triaged

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jan 6, 2024

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Jan 6, 2024

cjgillot added the A-mir-opt Area: MIR optimizations label Jan 6, 2024

This comment has been minimized.

Sign in to view

rustbot added perf-regression Performance regression. and removed S-waiting-on-perf Status: Waiting on a perf run to be completed. labels Jan 6, 2024

cjgillot force-pushed the dse-sandwich branch from b0c3689 to 366e4d3 Compare January 9, 2024 22:31

cjgillot added 3 commits January 11, 2024 09:58

Sandwich MIR optimizations between DSE.

0aedd6e

Bless ui-fulldeps.

974b727

Do not run simplify_locals inside DSE.

bc35ee4

The full pass is run short after.

cjgillot force-pushed the dse-sandwich branch from 366e4d3 to bc35ee4 Compare January 11, 2024 10:07

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Jan 12, 2024

This comment has been minimized.

Sign in to view

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Jan 13, 2024

rustbot assigned wesleywiser Jan 13, 2024

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jan 15, 2024

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Jan 16, 2024

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jan 16, 2024

bors added the merged-by-bors This PR was explicitly merged by bors. label Jan 16, 2024

bors merged commit fa0dc20 into rust-lang:master Jan 16, 2024
12 checks passed

rustbot added this to the 1.77.0 milestone Jan 16, 2024

rustbot added the perf-regression-triaged The performance regression has been triaged. label Jan 16, 2024

cjgillot deleted the dse-sandwich branch January 16, 2024 22:18

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Sandwich MIR optimizations between DSE. #119672

Sandwich MIR optimizations between DSE. #119672

cjgillot commented Jan 6, 2024

cjgillot commented Jan 6, 2024

This comment has been minimized.

bors commented Jan 6, 2024

This comment has been minimized.

bors commented Jan 6, 2024

This comment has been minimized.

rust-timer commented Jan 6, 2024

bors commented Jan 9, 2024

bors commented Jan 11, 2024

cjgillot commented Jan 12, 2024

This comment has been minimized.

bors commented Jan 12, 2024

bors commented Jan 12, 2024

This comment has been minimized.

rust-timer commented Jan 13, 2024

rustbot commented Jan 13, 2024

cjgillot commented Jan 13, 2024

oli-obk commented Jan 15, 2024

bors commented Jan 15, 2024

bors commented Jan 16, 2024

bors commented Jan 16, 2024

rust-log-analyzer commented Jan 16, 2024

oli-obk commented Jan 16, 2024

bors commented Jan 16, 2024

bors commented Jan 16, 2024

rust-timer commented Jan 16, 2024

lqd commented Jan 16, 2024

Sandwich MIR optimizations between DSE. #119672

Sandwich MIR optimizations between DSE. #119672

Conversation

cjgillot commented Jan 6, 2024

cjgillot commented Jan 6, 2024

This comment has been minimized.

bors commented Jan 6, 2024

This comment has been minimized.

bors commented Jan 6, 2024

This comment has been minimized.

rust-timer commented Jan 6, 2024

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

bors commented Jan 9, 2024

bors commented Jan 11, 2024

cjgillot commented Jan 12, 2024

This comment has been minimized.

bors commented Jan 12, 2024

bors commented Jan 12, 2024

This comment has been minimized.

rust-timer commented Jan 13, 2024

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

rustbot commented Jan 13, 2024

cjgillot commented Jan 13, 2024

oli-obk commented Jan 15, 2024

bors commented Jan 15, 2024

bors commented Jan 16, 2024

bors commented Jan 16, 2024

rust-log-analyzer commented Jan 16, 2024

oli-obk commented Jan 16, 2024

bors commented Jan 16, 2024

bors commented Jan 16, 2024

rust-timer commented Jan 16, 2024

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

lqd commented Jan 16, 2024