Update to LLVM 20 #135763

nikic · 2025-01-20T11:16:15Z

LLVM 20 GA is scheduled for March 11th. Rust 1.87 will be stable on May 15th.

Tested: host-x86_64, host-aarch64, apple, mingw, msvc

nikic · 2025-01-20T11:17:22Z

@bors try @rust-timer queue

bors · 2025-01-20T11:18:33Z

⌛ Trying commit fc35e8c with merge 79efb5f...

Update to LLVM 20 r? `@ghost`

bors · 2025-01-20T13:44:21Z

☀️ Try build successful - checks-actions
Build commit: 79efb5f (79efb5ff182537ce584a422f953ff24ab8f741db)

rust-timer · 2025-01-20T16:05:37Z

Finished benchmarking commit (79efb5f): comparison URL.

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

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 the most reliable metric that we have; it was used to determine the overall result at the top of this comment. However, even this metric can sometimes exhibit noise.

	mean	range	count
Regressions ❌ (primary)	5.2%	[0.3%, 34.9%]	8
Regressions ❌ (secondary)	0.6%	[0.3%, 1.1%]	14
Improvements ✅ (primary)	-1.3%	[-4.2%, -0.2%]	250
Improvements ✅ (secondary)	-1.0%	[-5.0%, -0.2%]	192
All ❌✅ (primary)	-1.1%	[-4.2%, 34.9%]	258

Max RSS (memory usage)

Results (primary -1.5%, secondary -1.3%)

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)	1.1%	[1.1%, 1.2%]	2
Regressions ❌ (secondary)	2.1%	[1.1%, 3.6%]	3
Improvements ✅ (primary)	-1.9%	[-3.8%, -1.0%]	11
Improvements ✅ (secondary)	-2.1%	[-3.4%, -0.9%]	13
All ❌✅ (primary)	-1.5%	[-3.8%, 1.2%]	13

Cycles

Results (primary -1.8%, secondary -5.0%)

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)	39.1%	[39.1%, 39.1%]	1
Regressions ❌ (secondary)	3.6%	[3.6%, 3.6%]	1
Improvements ✅ (primary)	-2.4%	[-4.2%, -1.1%]	73
Improvements ✅ (secondary)	-5.2%	[-13.6%, -0.9%]	36
All ❌✅ (primary)	-1.8%	[-4.2%, 39.1%]	74

Binary size

Results (primary -1.2%, secondary -3.0%)

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)	0.2%	[0.1%, 0.5%]	12
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.4%	[-7.1%, -0.0%]	96
Improvements ✅ (secondary)	-3.0%	[-7.2%, -0.1%]	90
All ❌✅ (primary)	-1.2%	[-7.1%, 0.5%]	108

Bootstrap: 765.63s -> 749.726s (-2.08%)
Artifact size: 326.02 MiB -> 335.29 MiB (2.85%)

nikic · 2025-01-27T16:18:17Z

The compiler-rt build (via compiler-builtins) for aarch64-unknown-uefi on test-various fails with:

warning: compiler_builtins@0.1.143: In file included from /checkout/src/llvm-project/compiler-rt/lib/builtins/cpu_model/aarch64.c:52:
warning: compiler_builtins@0.1.143: /checkout/src/llvm-project/compiler-rt/lib/builtins/cpu_model/aarch64/lse_atomics/windows.inc:2:10: fatal error: 'windows.h' file not found                                                                 
warning: compiler_builtins@0.1.143: #include <windows.h>
warning: compiler_builtins@0.1.143:          ^~~~~~~~~~~
warning: compiler_builtins@0.1.143: 1 error generated.
error: failed to run custom build command for `compiler_builtins v0.1.143`

The compilation command is:

  error occurred: Command "sccache" "clang-11" "-O3" "-ffunction-sections" "-fdata-sections" "--target=aarch64-unknown-windows-gnu" "-I" "/checkout/src/llvm-project/compiler-rt/lib/builtins" "-ffunction-sections" "-fdata-sections" "--target=aarch64-unknown-windows-gnu" "-fno-builtin" "-fvisibility=hidden" "-ffreestanding" "-I/checkout/obj/build/x86_64-unknown-linux-gnu/stage2-std/aarch64-unknown-uefi/release/build/compiler_builtins-de67acc96632c56e/out/include" "-fomit-frame-pointer" "-ffile-prefix-map=/checkout/src/llvm-project/compiler-rt=." "-DCOMPILER_RT_HAS_FLOAT16" "-DVISIBILITY_HIDDEN" "-o" "/checkout/obj/build/x86_64-unknown-linux-gnu/stage2-std/aarch64-unknown-uefi/release/build/compiler_builtins-de67acc96632c56e/out/670c1ed24249df13-aarch64.o" "-c" "/checkout/src/llvm-project/compiler-rt/lib/builtins/cpu_model/aarch64.c" with args clang-11 did not execute successfully (status code exit status: 1).

Related upstream change: llvm/llvm-project#116706

klensy · 2025-01-27T17:25:28Z

clang-11

What's wrong with version?

nikic · 2025-01-30T10:43:54Z

Testing host-aarch64.

@bors try

Update to LLVM 20 LLVM 20 GA is scheduled for March 11th. Rust 1.86 will be stable on April 3rd. * [x] rust-lang#135764 * [x] rust-lang#136134 * [x] rust-lang/compiler-builtins#752 * [ ] Update compiler-builtins Tested: arm-android, test-various, x86_64-rust-for-linux, x86_64-fuchsia, dist-various-1, dist-various-2, dist-s390x-linux, dist-riscv64-linux, dist-loongarch64-musl try-job: aarch64-gnu try-job: aarch64-gnu-debug try-job: dist-aarch64-linux

bors · 2025-01-30T10:45:06Z

⌛ Trying commit 881aa2d with merge 93cf5c1...

bors · 2025-01-30T14:02:32Z

☀️ Try build successful - checks-actions
Build commit: 93cf5c1 (93cf5c1ddb32e8f3a9fcdd8cd2cf78079ed6c78a)

nikic · 2025-01-30T14:11:49Z

Testing macos.

@bors try

Update to LLVM 20 LLVM 20 GA is scheduled for March 11th. Rust 1.86 will be stable on April 3rd. * [x] rust-lang#135764 * [x] rust-lang#136134 * [x] rust-lang/compiler-builtins#752 * [ ] Update compiler-builtins Tested: arm-android, test-various, x86_64-rust-for-linux, x86_64-fuchsia, dist-various-1, dist-various-2, dist-s390x-linux, dist-riscv64-linux, dist-loongarch64-musl try-job: dist-x86_64-apple try-job: dist-apple-various try-job: x86_64-apple-1 try-job: x86_64-apple-2 try-job: dist-aarch64-apple try-job: aarch64-apple

bors · 2025-01-30T14:12:59Z

⌛ Trying commit 881aa2d with merge 92ffaa8...

bors · 2025-01-30T18:08:52Z

☀️ Try build successful - checks-actions
Build commit: 92ffaa8 (92ffaa8cf69eca51dee2346f614ba8e990f37f34)

nikic · 2025-01-30T19:40:48Z

Testing mingw.

@bors try

bors · 2025-01-30T19:41:58Z

⌛ Trying commit 881aa2d with merge 6ff8ac7...

Update to LLVM 20 LLVM 20 GA is scheduled for March 11th. Rust 1.86 will be stable on April 3rd. * [x] rust-lang#135764 * [x] rust-lang#136134 * [x] rust-lang/compiler-builtins#752 * [ ] Update compiler-builtins Tested: arm-android, test-various, x86_64-rust-for-linux, x86_64-fuchsia, dist-various-1, dist-various-2, dist-s390x-linux, dist-riscv64-linux, dist-loongarch64-musl try-job: i686-mingw try-job: x86_64-mingw-1 try-job: x86_64-mingw-2 try-job: dist-i686-mingw try-job: dist-x86_64-mingw

nikic · 2025-01-31T14:06:25Z

I don't think windows-2025 is going to make a difference here. I'm seeing max-rss of ~2G on that TableGen invocation locally, so it does seem plausible that this could go OOM on i686, especially if parallelism is involved.

Testing a patch to cut memory usage down a bit, though it's only by about 100MB...

@bors retry

marcoieni · 2025-01-31T14:14:57Z

oh right, because i686 can only use a maximum of 4GB RAM. 💡
More reasons to demote this target maybe? 😅 I'll leave it to you to figure out 👍

nikic · 2025-01-31T15:21:52Z

@bors try

Update to LLVM 20 LLVM 20 GA is scheduled for March 11th. Rust 1.86 will be stable on April 3rd. * [x] rust-lang#135764 * [x] rust-lang#136134 * [x] rust-lang/compiler-builtins#752 * [ ] Update compiler-builtins Tested: host-x86_64, host-aarch64 try-job: i686-mingw try-job: x86_64-mingw-1 try-job: x86_64-mingw-2 try-job: dist-i686-mingw try-job: dist-x86_64-mingw

bors · 2025-01-31T15:23:06Z

⌛ Trying commit 9539a55 with merge a093dde...

mati865 · 2025-01-31T16:07:30Z

oh right, because i686 can only use a maximum of 4GB RAM. 💡
More reasons to demote this target maybe? 😅 I'll leave it to you to figure out 👍

Actually, it's half that size unless the program is linked with a special flag (the codebase must also support this). So, with 2 GiB Nikita has observed, we are right on the edge.

The same issue can manifest on any 32-bit Windows system (including those using MSVC). Note that it is LLVM's tblgen that fails, not GCC or ld.bfd.

LLVM generally does not cope well with 32-bit hosts, but so far, the problem has mostly been due to mmap having insufficient address space.

One solution is cross-compiling with a 64-bit toolchain, but that requires changes to the entire build process.

bors · 2025-01-31T19:05:11Z

☀️ Try build successful - checks-actions
Build commit: a093dde (a093dde8bafff0c2c30e443df480946ec24196fe)

nikic · 2025-01-31T20:50:18Z

Looks like that was enough. Now testing msvc...

@bors try

bors · 2025-01-31T20:51:27Z

⌛ Trying commit 9539a55 with merge 23c742f...

Update to LLVM 20 LLVM 20 GA is scheduled for March 11th. Rust 1.86 will be stable on April 3rd. * [x] rust-lang#135764 * [x] rust-lang#136134 * [x] rust-lang/compiler-builtins#752 * [ ] llvm/llvm-project#125287 * [ ] Update compiler-builtins Tested: host-x86_64, host-aarch64, mingw try-job: x86_64-msvc-1 try-job: x86_64-msvc-2 try-job: i686-msvc-1 try-job: i686-msvc-2 try-job: dist-x86_64-msvc try-job: dist-i686-msvc try-job: dist-aarch64-msvc

bors · 2025-02-01T00:16:27Z

💔 Test failed - checks-actions

nikic · 2025-02-01T09:16:14Z

Looks like the embedded command line arguments in https://github.com/rust-lang/rust/blob/master/tests/run-make/pdb-buildinfo-cl-cmd/filecheck.txt are no longer quoted. I think this is because llvm/llvm-project@e190d07 moved the command line printing to the frontend, but the Rust adjustment in ad0eceb just prints the arguments directly, instead of going through llvm::sys::printArg.

matthiaskrgr · 2025-02-01T09:50:19Z

@bors r-

note that bors retry means "please try to merge again (as in, after approval, CI failed spuriously) and ´bory try´ means "please run a try job" so "retry" is not a "please re-try the failed try" job! 😅

the8472 · 2025-02-03T10:15:27Z

Actually, it's half that size unless the program is linked with a special flag (the codebase must also support this). So, with 2 GiB Nikita has observed, we are right on the edge.

The flag can also be set after linking via editbin /LARGEADDRESSAWARE <executable>. Might be worth a try, programs doing pointer shenanigans aren't that common.

The formatting of the command line arguments has been moved to the frontend in: llvm/llvm-project@e190d07 However, the Rust logic introduced in rust-lang@ad0eceb did not replicate the previous argument quoting behavior.

nikic · 2025-02-03T15:20:04Z

@bors try

Update to LLVM 20 LLVM 20 GA is scheduled for March 11th. Rust 1.86 will be stable on April 3rd. * [x] rust-lang#135764 * [x] rust-lang#136134 * [x] rust-lang/compiler-builtins#752 * [ ] llvm/llvm-project#125287 * [ ] Update compiler-builtins (blocked on rust-lang#135501) Tested: host-x86_64, host-aarch64, apple, mingw try-job: x86_64-msvc-1 try-job: i686-msvc-1 try-job: dist-aarch64-msvc try-job: x86_64-msvc-ext1 try-job: x86_64-msvc-ext2 try-job: x86_64-msvc-ext3

bors · 2025-02-03T15:21:18Z

⌛ Trying commit ee8284e with merge ad23dfe...

mati865 · 2025-02-03T17:23:27Z

The flag can also be set after linking via editbin /LARGEADDRESSAWARE <executable>. Might be worth a try, programs doing pointer shenanigans aren't that common.

Editbin is not the most manageable approach because you would have to modify LLVM's build system (notice it failed halfway through the build), but you can append linker flags to all Windows platforms.

I am not familiar enough with LLVM code to judge whether something would break, but yes, that might be another option to try.

bors · 2025-02-03T18:01:42Z

☀️ Try build successful - checks-actions
Build commit: ad23dfe (ad23dfe0eb9153bd71988323086e143f7b3947f6)

tgross35 · 2025-02-06T01:12:27Z

* [ ]  Update compiler-builtins (blocked on [Inject `compiler_builtins` during postprocessing and ensure it is made private #135501](https://github.com/rust-lang/rust/pull/135501))

I wound up just doing the revert to unblock this, should be done as of #136537

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 20, 2025

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 20, 2025

bors added a commit to rust-lang-ci/rust that referenced this pull request Jan 20, 2025

Auto merge of rust-lang#135763 - nikic:llvm-20, r=<try>

79efb5f

Update to LLVM 20 r? `@ghost`

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 20, 2025

nikic force-pushed the llvm-20 branch from fc35e8c to 110bee6 Compare January 27, 2025 14:14

This comment has been minimized.

Sign in to view

nikic force-pushed the llvm-20 branch from 110bee6 to da0ec57 Compare January 27, 2025 14:20

nikic mentioned this pull request Jan 28, 2025

Don't build out of line atomics support code for uefi rust-lang/compiler-builtins#752

Merged

nikic force-pushed the llvm-20 branch from 3131edb to 881aa2d Compare January 30, 2025 10:33

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-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Jan 31, 2025

This comment has been minimized.

Sign in to view

bors added the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Feb 1, 2025

bors removed the S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. label Feb 1, 2025

tones111 mentioned this pull request Feb 2, 2025

Upgrade rust toolchain to nightly-2025-01-01 Rahix/avr-hal#585

Open

Quote embedded codeview command line arguments

ee8284e

The formatting of the command line arguments has been moved to the frontend in: llvm/llvm-project@e190d07 However, the Rust logic introduced in rust-lang@ad0eceb did not replicate the previous argument quoting behavior.

dtcxzyw mentioned this pull request Feb 3, 2025

Bump clang/rustc frontend version dtcxzyw/llvm-opt-benchmark#2077

Open

kjetilkjeka mentioned this pull request Feb 4, 2025

NVPTX: Debug section symbol naming in .ptx output #99248

Open

the8472 mentioned this pull request Feb 9, 2025

avoid phi node for pointers flowing into Vec appends #130998

Draft

Update to LLVM 20 #135763

Are you sure you want to change the base?

Update to LLVM 20 #135763

Conversation

nikic commented Jan 20, 2025 • edited Loading

nikic commented Jan 20, 2025

This comment has been minimized.

bors commented Jan 20, 2025

bors commented Jan 20, 2025

This comment has been minimized.

rust-timer commented Jan 20, 2025

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

Instruction count

Max RSS (memory usage)

Cycles

Binary size

This comment has been minimized.

nikic commented Jan 27, 2025 • edited Loading

klensy commented Jan 27, 2025

nikic commented Jan 30, 2025

bors commented Jan 30, 2025

bors commented Jan 30, 2025

nikic commented Jan 30, 2025

bors commented Jan 30, 2025

bors commented Jan 30, 2025

nikic commented Jan 30, 2025

bors commented Jan 30, 2025

nikic commented Jan 31, 2025

marcoieni commented Jan 31, 2025

nikic commented Jan 31, 2025

bors commented Jan 31, 2025

mati865 commented Jan 31, 2025

bors commented Jan 31, 2025

nikic commented Jan 31, 2025

bors commented Jan 31, 2025

This comment has been minimized.

bors commented Feb 1, 2025

nikic commented Feb 1, 2025 • edited Loading

matthiaskrgr commented Feb 1, 2025 • edited Loading

the8472 commented Feb 3, 2025

nikic commented Feb 3, 2025

bors commented Feb 3, 2025

mati865 commented Feb 3, 2025

bors commented Feb 3, 2025

tgross35 commented Feb 6, 2025

nikic commented Jan 20, 2025 •

edited

Loading

nikic commented Jan 27, 2025 •

edited

Loading

nikic commented Feb 1, 2025 •

edited

Loading

matthiaskrgr commented Feb 1, 2025 •

edited

Loading