[WIP] Build `rustc` with a single CGU on x64 Linux #107651

Kobzol · 2023-02-03T23:05:57Z

Follow-up attempt to #87650. I wonder if anything changed with the addition of LTO.

I also enabled a single CGU only for the actual build of the compiler on CI, so that we can better see the perf. effects on the bootstrap benchmark.

rustbot · 2023-02-03T23:06:03Z

r? @pietroalbini

(rustbot has picked a reviewer for you, use r? to override)

Kobzol · 2023-02-03T23:06:05Z

@bors try @rust-timer queue

bors · 2023-02-03T23:06:13Z

⌛ Trying commit 96c912abfc4b44ce5f8fca252500880ffba28d3e with merge 2a203d5ea41b38173f8e09d32d989d0c26770e28...

bors · 2023-02-04T02:23:11Z

☀️ Try build successful - checks-actions
Build commit: 2a203d5ea41b38173f8e09d32d989d0c26770e28 (2a203d5ea41b38173f8e09d32d989d0c26770e28)

rust-timer · 2023-02-04T06:08:50Z

Finished benchmarking commit (2a203d5ea41b38173f8e09d32d989d0c26770e28): 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)	0.9%	[0.2%, 3.1%]	68
Regressions ❌ (secondary)	4.7%	[0.5%, 27.2%]	57
Improvements ✅ (primary)	-1.1%	[-4.1%, -0.2%]	71
Improvements ✅ (secondary)	-1.7%	[-7.6%, -0.5%]	121
All ❌✅ (primary)	-0.1%	[-4.1%, 3.1%]	139

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)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.1%	[-6.4%, -0.5%]	86
Improvements ✅ (secondary)	-3.5%	[-7.7%, -0.9%]	153
All ❌✅ (primary)	-2.1%	[-6.4%, -0.5%]	86

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)	2.0%	[1.4%, 2.9%]	3
Regressions ❌ (secondary)	7.2%	[2.1%, 17.5%]	20
Improvements ✅ (primary)	-2.3%	[-5.4%, -0.8%]	52
Improvements ✅ (secondary)	-3.1%	[-4.7%, -1.1%]	77
All ❌✅ (primary)	-2.1%	[-5.4%, 2.9%]	55

Zoxc · 2023-02-04T07:01:27Z

This seems to perform similarly to #107560, which make sense since that compiles most crates with 1 CGU. It's an overall improvement, but with some regressions. It does seem like there's some problem with ThinLTO + PGO for match-stress, but that is a stress test and not terribly important.

Kobzol · 2023-02-04T08:27:42Z

The CI time hit is also not that terrible, 8600s (I saw a recent master commit had 8000s), not sure what's the average recently (@Mark-Simulacrum ).

We can get that time down, too. LLVM rebuild time can be lowered, and we can also speedup rustc builds - e.g. we don't need to build with 1 CGU when we do LLVM PGO/BOLT, and I think that we can also skip LTO/1 CGU when building stage 1 (we'd need to modify bootstrap though).

Zoxc · 2023-02-23T22:47:17Z

Building with 1 CGU would be useful to make the --emit options better match the CI builds too.

Zoxc · 2023-04-07T13:21:59Z

I did some local benchmarks to test 16 CGUs (Before) vs. 1 CGU (After) with ThinLTO:

Benchmark	Before	After
Benchmark	Time	Time	%
🟣 clap:check:unchanged	0.4631s	0.4475s	💚 -3.37%
🟣 hyper:check:unchanged	0.1528s	0.1472s	💚 -3.65%
🟣 regex:check:unchanged	0.3433s	0.3370s	💚 -1.83%
🟣 syn:check:unchanged	0.6308s	0.6065s	💚 -3.84%
🟣 syntex_syntax:check:unchanged	1.8294s	1.7532s	💚 -4.16%
Total	3.4193s	3.2914s	💚 -3.74%
Summary	1.0000s	0.9663s	💚 -3.37%

Benchmark	Before	After
Benchmark	Time	Time	%
🟣 clap:check	1.6721s	1.5579s	💚 -6.83%
🟣 hyper:check	0.2592s	0.2418s	💚 -6.69%
🟣 regex:check	0.9330s	0.8702s	💚 -6.74%
🟣 syn:check	1.4963s	1.3939s	💚 -6.85%
🟣 syntex_syntax:check	5.8074s	5.3920s	💚 -7.15%
Total	10.1681s	9.4558s	💚 -7.00%
Summary	1.0000s	0.9315s	💚 -6.85%

Kobzol · 2023-05-17T08:02:27Z

@bors try @rust-timer queue

bors · 2023-05-17T08:02:40Z

⌛ Trying commit 96c912abfc4b44ce5f8fca252500880ffba28d3e with merge 0b5019f4e7c7c5e8190c0d4e9686840a7a3cfc00...

bors · 2023-05-17T10:25:44Z

💔 Test failed - checks-actions

Kobzol · 2023-05-25T09:54:50Z

@bors try @rust-timer queue

bors · 2023-09-05T00:02:03Z

☀️ Try build successful - checks-actions
Build commit: 146169d35b066f541a24c7489cd7ef4397300cf5 (146169d35b066f541a24c7489cd7ef4397300cf5)

rust-timer · 2023-09-05T01:25:47Z

Finished benchmarking commit (146169d35b066f541a24c7489cd7ef4397300cf5): 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%	[0.3%, 2.0%]	31
Regressions ❌ (secondary)	1.7%	[0.6%, 2.4%]	16
Improvements ✅ (primary)	-1.3%	[-4.1%, -0.3%]	96
Improvements ✅ (secondary)	-1.7%	[-3.7%, -0.5%]	122
All ❌✅ (primary)	-0.8%	[-4.1%, 2.0%]	127

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)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.1%	[-4.9%, -0.6%]	68
Improvements ✅ (secondary)	-3.1%	[-7.6%, -0.5%]	136
All ❌✅ (primary)	-2.1%	[-4.9%, -0.6%]	68

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)	-	-	0
Regressions ❌ (secondary)	2.6%	[2.5%, 2.6%]	2
Improvements ✅ (primary)	-2.5%	[-7.1%, -0.7%]	72
Improvements ✅ (secondary)	-3.8%	[-15.6%, -0.8%]	99
All ❌✅ (primary)	-2.5%	[-7.1%, -0.7%]	72

Binary size

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

Bootstrap: 629.118s -> 625.733s (-0.54%)
Artifact size: 316.16 MiB -> 271.45 MiB (-14.14%)

Kobzol · 2023-09-15T14:22:10Z

@bors try

Trying with LLVM 16.

bors · 2023-09-15T14:22:21Z

⌛ Trying commit a47a9eb with merge 3cdcb31...

[WIP] Build `rustc` with a single CGU on x64 Linux Follow-up attempt to rust-lang#87650. I wonder if anything changed with the addition of LTO. I also enabled a single CGU only for the actual build of the compiler on CI, so that we can better see the perf. effects on the bootstrap benchmark.

Kobzol · 2023-09-15T14:53:59Z

@bors try

bors · 2023-09-15T14:54:08Z

⌛ Trying commit dba68fa with merge a3b1fd2...

[WIP] Build `rustc` with a single CGU on x64 Linux Follow-up attempt to rust-lang#87650. I wonder if anything changed with the addition of LTO. I also enabled a single CGU only for the actual build of the compiler on CI, so that we can better see the perf. effects on the bootstrap benchmark.

bors · 2023-09-15T15:00:52Z

💔 Test failed - checks-actions

bors · 2023-09-20T04:10:11Z

☔ The latest upstream changes (presumably #115959) made this pull request unmergeable. Please resolve the merge conflicts.

rust-log-analyzer · 2023-09-28T19:38:50Z

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

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

Prepare all required actions
Getting action download info
Download action repository 'actions/checkout@v4' (SHA:8ade135a41bc03ea155e62e844d188df1ea18608)
Download action repository 'actions/upload-artifact@v3' (SHA:a8a3f3ad30e3422c9c7b888a15615d19a852ae32)
Complete job name: PR - mingw-check-tidy
git config --global core.autocrlf false
shell: /usr/bin/bash --noprofile --norc -e -o pipefail {0}
---
GITHUB_ENV=/home/runner/work/_temp/_runner_file_commands/set_env_ee7d33b2-1f86-4f3c-a858-bb1657e6f732
GITHUB_EVENT_NAME=pull_request
GITHUB_EVENT_PATH=/home/runner/work/_temp/_github_workflow/event.json
GITHUB_GRAPHQL_URL=https://api.github.com/graphql
GITHUB_HEAD_REF=ci-single-cgu
GITHUB_JOB=pr
GITHUB_PATH=/home/runner/work/_temp/_runner_file_commands/add_path_ee7d33b2-1f86-4f3c-a858-bb1657e6f732
GITHUB_REF=refs/pull/107651/merge
GITHUB_REF_NAME=107651/merge
GITHUB_REF_PROTECTED=false
---

Sending build context to Docker daemon  432.1kB

Step 1/10 : FROM ubuntu:22.04
Head "https://registry-1.docker.io/v2/library/ubuntu/manifests/22.04": Get "https://auth.docker.io/token?account=githubactions&scope=repository%3Alibrary%2Fubuntu%3Apull&service=registry.docker.io": EOF
DEPRECATED: The legacy builder is deprecated and will be removed in a future release.
            BuildKit is currently disabled; enable it by removing the DOCKER_BUILDKIT=0
            environment-variable.


Sending build context to Docker daemon  432.1kB

Step 1/10 : FROM ubuntu:22.04
Get "https://registry-1.docker.io/v2/library/ubuntu/manifests/sha256:aabed3296a3d45cede1dc866a24476c4d7e093aa806263c27ddaadbdce3c1054": EOF
DEPRECATED: The legacy builder is deprecated and will be removed in a future release.
            BuildKit is currently disabled; enable it by removing the DOCKER_BUILDKIT=0
            environment-variable.


Sending build context to Docker daemon  432.1kB

Step 1/10 : FROM ubuntu:22.04
Head "https://registry-1.docker.io/v2/library/ubuntu/manifests/22.04": EOF
DEPRECATED: The legacy builder is deprecated and will be removed in a future release.
            BuildKit is currently disabled; enable it by removing the DOCKER_BUILDKIT=0
            environment-variable.


Sending build context to Docker daemon  432.1kB

Step 1/10 : FROM ubuntu:22.04
Get "https://registry-1.docker.io/v2/library/ubuntu/manifests/sha256:aabed3296a3d45cede1dc866a24476c4d7e093aa806263c27ddaadbdce3c1054": EOF
##[error]Process completed with exit code 1.
Post job cleanup.

lnicola · 2023-10-01T15:02:32Z

How did this get merged by bors? There's no r+.

lqd · 2023-10-01T15:14:20Z

This draft PR contains a single commit ca59652 which was merged in #115554

rustbot assigned pietroalbini Feb 3, 2023

rustbot added A-testsuite Area: The testsuite used to check the correctness of rustc S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-infra Relevant to the infrastructure team, which will review and decide on the PR/issue. labels Feb 3, 2023