Update to LLVM 22 #150722

nikic · 2026-01-06T11:26:00Z

Scheduled release date: Feb 24
1.94 becomes stable: Mar 5

Depends on:

Update Illumos sysroot #150756
[PowerPC] Branch target out of range llvm/llvm-project#175190
perf regressions

rustbot · 2026-01-06T11:26:03Z

These commits modify compiler targets.
(See the Target Tier Policy.)

rustbot · 2026-01-06T11:26:05Z

⚠️ Warning ⚠️

Some commits in this PR modify submodules.

If this was not intentional, see I changed a submodule on accident in the rustc dev guide.

nikic · 2026-01-06T11:39:10Z

@bors try @rust-timer queue

Update to LLVM 22

rust-bors · 2026-01-06T14:36:15Z

☀️ Try build successful (CI)
Build commit: dabe9cd (dabe9cd2e5a1630ac5ed5f1b87f891b6dc563025, parent: da476f1942868cdf94ed88b01ea31170cfe95047)

rust-timer · 2026-01-06T15:17:47Z

Finished benchmarking commit (dabe9cd): 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)	0.9%	[0.2%, 8.9%]	228
Regressions ❌ (secondary)	1.4%	[0.2%, 8.0%]	260
Improvements ✅ (primary)	-0.8%	[-1.1%, -0.3%]	6
Improvements ✅ (secondary)	-2.2%	[-8.0%, -0.2%]	28
All ❌✅ (primary)	0.9%	[-1.1%, 8.9%]	234

Max RSS (memory usage)

Results (primary 1.0%, 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)	1.0%	[0.8%, 1.3%]	2
Regressions ❌ (secondary)	1.8%	[0.9%, 2.4%]	8
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.7%	[-4.8%, -2.6%]	10
All ❌✅ (primary)	1.0%	[0.8%, 1.3%]	2

Cycles

Results (primary 3.2%, secondary 2.1%)

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

	mean	range	count
Regressions ❌ (primary)	3.2%	[1.6%, 10.2%]	19
Regressions ❌ (secondary)	3.3%	[1.3%, 7.4%]	38
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-4.8%	[-7.5%, -2.5%]	7
All ❌✅ (primary)	3.2%	[1.6%, 10.2%]	19

Binary size

Results (primary -0.6%, secondary -1.1%)

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

	mean	range	count
Regressions ❌ (primary)	0.1%	[0.1%, 0.1%]	4
Regressions ❌ (secondary)	2.6%	[0.1%, 5.1%]	2
Improvements ✅ (primary)	-0.8%	[-1.5%, -0.1%]	14
Improvements ✅ (secondary)	-1.1%	[-3.9%, -0.0%]	93
All ❌✅ (primary)	-0.6%	[-1.5%, 0.1%]	18

Bootstrap: 473.133s -> 480.877s (1.64%)
Artifact size: 390.77 MiB -> 402.18 MiB (2.92%)

rust-bors · 2026-01-06T18:45:25Z

☔ The latest upstream changes made this pull request unmergeable. Please resolve the merge conflicts.

bors · 2026-01-06T19:43:13Z

☔ The latest upstream changes (presumably #150726) made this pull request unmergeable. Please resolve the merge conflicts.

nikic · 2026-01-08T11:49:52Z

Based on helloworld, the perf regressions seem to be related to the allocator somehow. Previously tcache_alloc_small_hard was called 3 times, now it's called 189 times. The total number of allocations is smaller, but the time spent in the allocator is larger.

nikic · 2026-01-08T14:29:53Z

Or maybe the issue is not actually the allocator behavior itself. I suspect that we might have lost LTO on jemalloc and tcache_alloc_small_hard previously got inlined into malloc_default but now no longer is. Possibly updating the host toolchain at the same time so that the versions match will help.

This changed in: llvm/llvm-project@853760b

Since LLVM 22 the alloca must be passed directly. Do this by stripping the addrspacecast if it exists.

rustbot · 2026-01-08T15:01:13Z

This PR was rebased onto a different main 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.

nikic · 2026-01-08T15:02:14Z

@bors try @rust-timer queue

Update to LLVM 22

rust-bors · 2026-01-08T18:36:46Z

☀️ Try build successful (CI)
Build commit: b02fa67 (b02fa67ce3a0b9c29b6368e7afc02a38bdadff30, parent: 32fe406b5e71afbb0d8b95280e50e67d1549224c)

rust-timer · 2026-01-08T19:16:03Z

Finished benchmarking commit (b02fa67): comparison URL.

Overall result: ❌✅ regressions and improvements - BENCHMARK(S) FAILED

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

❗ ❗ ❗ ❗ ❗
Warning ⚠️: The following benchmark(s) failed to build:

rustc
Job failure

❗ ❗ ❗ ❗ ❗

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)	4.1%	[0.8%, 7.4%]	2
Regressions ❌ (secondary)	2.8%	[0.4%, 6.2%]	6
Improvements ✅ (primary)	-0.6%	[-2.6%, -0.2%]	233
Improvements ✅ (secondary)	-0.8%	[-10.0%, -0.1%]	266
All ❌✅ (primary)	-0.6%	[-2.6%, 7.4%]	235

Max RSS (memory usage)

Results (primary 2.5%, 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)	2.5%	[2.2%, 2.8%]	2
Regressions ❌ (secondary)	2.0%	[0.6%, 3.7%]	27
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-4.1%	[-5.6%, -2.1%]	9
All ❌✅ (primary)	2.5%	[2.2%, 2.8%]	2

Cycles

Results (primary 4.1%, 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)	4.1%	[2.4%, 8.3%]	4
Regressions ❌ (secondary)	3.1%	[2.1%, 7.4%]	12
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-4.7%	[-7.8%, -2.3%]	8
All ❌✅ (primary)	4.1%	[2.4%, 8.3%]	4

Binary size

Results (primary -0.5%, secondary -1.1%)

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

	mean	range	count
Regressions ❌ (primary)	0.1%	[0.1%, 0.1%]	4
Regressions ❌ (secondary)	2.6%	[0.1%, 5.1%]	2
Improvements ✅ (primary)	-0.7%	[-1.5%, -0.1%]	24
Improvements ✅ (secondary)	-1.2%	[-3.8%, -0.0%]	93
All ❌✅ (primary)	-0.5%	[-1.5%, 0.1%]	28

Bootstrap: missing data
Artifact size: 390.80 MiB -> 402.96 MiB (3.11%)

Kobzol · 2026-01-08T19:39:21Z

The rustc bootstrap failure is because the try build hasn't produced the rust-dev component with LLVM, as it is skipped in "fast" try builds. rustc-perf tried to download it because LLVM changed in this PR. I wonder how could this have worked in the past few months 😮

nikic · 2026-01-08T19:48:20Z

Given that the perf run from two days ago worked, maybe something related to the bors switchover affected this?

At least the across the board regressions are gone now. Looks like we need to update the host toolchain in sync going forward to avoid this issue.

Kobzol · 2026-01-08T22:04:59Z

Given that the perf run from two days ago worked, maybe something related to the bors switchover affected this?

Yeah, that's most likely. But the weird thing is that I actually consider the new behavior to be correct 😆 Anyway, I'll investigate.

As Intel has walked back on the existence of AVX 10.1-256, LLVM no longer uses evex512 and avx-10.n-512 are now avx-10.n instead, so we can skip all the special handling on LLVM 22.

Kobzol · 2026-01-09T14:16:56Z

Btw, looks like your LLVM changes @Mark-Simulacrum made BOLT optimizations faster 🎉! 5m => 3m for LLVM, and 10m => 5m for rustc.

This comment has been minimized.

Sign in to view

rust-bors bot added a commit that referenced this pull request Jan 6, 2026

Auto merge of #150722 - nikic:llvm-22, r=<try>

dabe9cd

Update to LLVM 22

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Jan 6, 2026

This comment has been minimized.

Sign in to view

This comment was marked as outdated.

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, 2026

This comment was marked as outdated.

Sign in to view

Kobzol mentioned this pull request Jan 6, 2026

Only require try permissions for marking a PR as rollup rust-lang/bors#553

Merged

nikic force-pushed the llvm-22 branch from b0ff5bb to 78019a6 Compare January 6, 2026 15:46

This comment has been minimized.

Sign in to view

nikic added 4 commits January 8, 2026 16:01

Update to LLVM 22

2b246f8

Update amdgpu data layout

2387f99

This changed in: llvm/llvm-project@853760b

Avoid passing addrspacecast to lifetime intrinsics

af8e1f5

Since LLVM 22 the alloca must be passed directly. Do this by stripping the addrspacecast if it exists.

Update host toolchain

aafd9df

nikic force-pushed the llvm-22 branch from 78019a6 to aafd9df Compare January 8, 2026 15:01

rustbot added A-CI Area: Our Github Actions CI A-testsuite Area: The testsuite used to check the correctness of rustc T-infra Relevant to the infrastructure team, which will review and decide on the PR/issue. labels Jan 8, 2026

This comment has been minimized.

Sign in to view

rust-bors bot added a commit that referenced this pull request Jan 8, 2026

Auto merge of #150722 - nikic:llvm-22, r=<try>

b02fa67

Update to LLVM 22

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Jan 8, 2026

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 8, 2026

Don't use evex512 with LLVM 22

7f2f573

As Intel has walked back on the existence of AVX 10.1-256, LLVM no longer uses evex512 and avx-10.n-512 are now avx-10.n instead, so we can skip all the special handling on LLVM 22.

Uh oh!

Update to LLVM 22 #150722

Are you sure you want to change the base?

Update to LLVM 22 #150722

Conversation

nikic commented Jan 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Jan 6, 2026

Uh oh!

rustbot commented Jan 6, 2026

Uh oh!

nikic commented Jan 6, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

rust-bors bot commented Jan 6, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Jan 6, 2026

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

Uh oh!

This comment was marked as outdated.

This comment was marked as outdated.

This comment has been minimized.

rust-bors bot commented Jan 6, 2026

Uh oh!

bors commented Jan 6, 2026

Uh oh!

nikic commented Jan 8, 2026

Uh oh!

nikic commented Jan 8, 2026

Uh oh!

rustbot commented Jan 8, 2026

Uh oh!

nikic commented Jan 8, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

rust-bors bot commented Jan 8, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Jan 8, 2026

Overall result: ❌✅ regressions and improvements - BENCHMARK(S) FAILED

Uh oh!

Kobzol commented Jan 8, 2026

Uh oh!

nikic commented Jan 8, 2026

Uh oh!

Kobzol commented Jan 8, 2026

Uh oh!

Kobzol commented Jan 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

nikic commented Jan 6, 2026 •

edited

Loading