Use LLVM easyblock for Clang 18 easyconfigs by Flamefire · Pull Request #23055 · easybuilders/easybuild-easyconfigs

Flamefire · 2025-06-06T11:26:58Z

Pulled out of #23028

Requires:

Don't install libomp aliases by default in LLVM easyblock easybuild-easyblocks#3825

Comparing the build with this and previous Clang easyblock shows that this fixes installing files for every single GPU architecture instead of only the intended ones, see easybuilders/easybuild-easyblocks#3755 (comment)

Also enabling testing as suggested in Slack and related issues.

easybuild/easyconfigs/c/Clang/Clang-18.1.8-GCCcore-13.3.0.eb

boegel · 2025-07-03T20:05:05Z

@boegelbot please test @ jsc-zen3-a100
CORE_CNT=16
EB_ARGS="--include-easyblocks-from-pr 3755"

boegelbot · 2025-07-03T20:10:09Z

@boegel: Request for testing this PR well received on jsczen3l1.int.jsc-zen3.fz-juelich.de

PR test command 'if [[ develop != 'develop' ]]; then EB_BRANCH=develop ./easybuild_develop.sh 2> /dev/null 1>&2; EB_PREFIX=/home/boegelbot/easybuild/develop source init_env_easybuild_develop.sh; fi; EB_PR=23055 EB_ARGS="--include-easyblocks-from-pr 3755" EB_CONTAINER= EB_REPO=easybuild-easyconfigs EB_BRANCH=develop /opt/software/slurm/bin/sbatch --job-name test_PR_23055 --ntasks="16" --partition=jsczen3g --gres=gpu:1 ~/boegelbot/eb_from_pr_upload_jsc-zen3.sh' executed!

exit code: 0
output:

Submitted batch job 7127

Test results coming soon (I hope)...

Details

- notification for comment with ID 3033481563 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

boegel

lgtm

boegelbot · 2025-07-04T00:08:20Z

Test report by @boegelbot
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#3755
FAILED
Build succeeded for 0 out of 2 (2 easyconfigs in total)
jsczen3g1.int.jsc-zen3.fz-juelich.de - Linux Rocky Linux 9.5, x86_64, AMD EPYC-Milan Processor (zen3), 1 x NVIDIA NVIDIA A100 80GB PCIe, 555.42.06, Python 3.9.21
See https://gist.github.com/boegelbot/4946794639009dea78fac0ef78666115 for a full test report.

boegel · 2025-07-04T01:59:56Z

Test report by @boegel
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#3755, easybuilders/easybuild-easyblocks#3746
FAILED
Build succeeded for 0 out of 2 (2 easyconfigs in total)
node3527.doduo.os - Linux RHEL 9.4, x86_64, AMD EPYC 7552 48-Core Processor (zen2), Python 3.9.18
See https://gist.github.com/boegel/5c779b5665424a098bff53c60b52b157 for a full test report.

Flamefire · 2025-07-04T10:43:37Z

@Crivella It does not. It simply downloads the easyblock as-is and includes the path to the downloaded location. So only one of them is used, which one depends on the order we add the new paths.

I don't think merging is even possible: How would you handle conflicts (when you download diffs) and building from a merged PR?

Crivella · 2025-07-04T10:58:45Z

Yeah I assumed so, but also deleted the comment as i noticed it was 2 separate ones clang/llvm and not llvm/llvm my bad there

Flamefire · 2025-07-04T11:44:44Z

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#3825
FAILED
Build succeeded for 0 out of 2 (2 easyconfigs in total)
n1266 - Linux RHEL 8.9 (Ootpa), x86_64, Intel(R) Xeon(R) Platinum 8470 (sapphirerapids), Python 3.9.18
See https://gist.github.com/Flamefire/9e4f8344863a69fef0c53b2d1bfa156a for a full test report.

Flamefire · 2025-07-04T13:21:38Z

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#3825
FAILED
Build succeeded for 0 out of 2 (2 easyconfigs in total)
i7184 - Linux Rocky Linux 8.9 (Green Obsidian), x86_64, AMD EPYC 7702 64-Core Processor (zen2), Python 3.9.18
See https://gist.github.com/Flamefire/822969cadc704adbe01f255db2851074 for a full test report.

Flamefire · 2025-07-05T09:38:49Z

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#3825
SUCCESS
Build succeeded for 2 out of 2 (2 easyconfigs in total)
n1711 - Linux RHEL 8.9 (Ootpa), x86_64, Intel(R) Xeon(R) Platinum 8470 (sapphirerapids), Python 3.9.18
See https://gist.github.com/Flamefire/9ff51b893436ba678293583f1ac06e4e for a full test report.

Flamefire · 2025-07-05T10:55:51Z

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#3825
FAILED
Build succeeded for 1 out of 2 (2 easyconfigs in total)
c106 - Linux AlmaLinux 9.4, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 560.35.03, Python 3.9.18
See https://gist.github.com/Flamefire/e01ce4db1580b17b2285d1c3c9c211ec for a full test report.

Flamefire · 2025-07-06T00:21:39Z

Test report by @Flamefire
Using easyblocks from PR(s) easybuilders/easybuild-easyblocks#3825
FAILED
Build succeeded for 0 out of 2 (2 easyconfigs in total)
i7007 - Linux Rocky Linux 8.9 (Green Obsidian), x86_64, AMD EPYC 7702 64-Core Processor (zen2), Python 3.9.18
See https://gist.github.com/Flamefire/273c8538a40140ef78b064f719679310 for a full test report.

Flamefire · 2025-07-09T13:55:29Z

The tests fail mostly with:

error while loading shared libraries: libatomic.so.1: cannot open shared object file: No such file or directory

# | : CommandLine Error: Option 'debug-counter' registered more than once!

The 2nd seems to be a known issue with Clang 18, the other is weird: Their test tool run.py starts in a clean environment without LD_LIBRARY_PATH.

LLVM 20 works. Looks like it is supposed to add an rpath, but can't find anything relevant in the logs.

Maybe we should keep skipping the tests?

Crivella · 2025-07-09T16:20:24Z

We are gonna need the same patches that are applied for LLVM to pass all the tests

Crivella · 2025-07-09T16:22:55Z

In particular for the libatomic error

https://github.com/easybuilders/easybuild-easyconfigs/blob/develop/easybuild/easyconfigs/l/LLVM/LLVM-18.1.8_envintest.patch

This is needed to ensure LD_LIBRARY_PATH is passed when building the tests where needed.
This is a problem i had already encountered when building LLVM on the BOT where libatomic is in a non standard location

See

[test suite] Running test-suite with non standard GCC's libatomic location llvm/llvm-project#109949

Flamefire · 2025-07-25T09:55:49Z

Test report by @Flamefire
SUCCESS
Build succeeded for 2 out of 2 (2 easyconfigs in total)
n1318 - Linux RHEL 8.9 (Ootpa), x86_64, Intel(R) Xeon(R) Platinum 8470 (sapphirerapids), Python 3.9.18
See https://gist.github.com/Flamefire/1506d8fb6d276c9456d52ef40478e9d3 for a full test report.

Flamefire · 2025-07-25T11:08:15Z

Test report by @Flamefire
SUCCESS
Build succeeded for 2 out of 2 (2 easyconfigs in total)
c144 - Linux AlmaLinux 9.4, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 560.35.03, Python 3.9.18
See https://gist.github.com/Flamefire/0333f41a22364944b71ac2a1a674512c for a full test report.

Flamefire · 2025-07-25T15:01:15Z

Test report by @Flamefire
SUCCESS
Build succeeded for 2 out of 2 (2 easyconfigs in total)
c144 - Linux AlmaLinux 9.4, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 560.35.03, Python 3.9.18
See https://gist.github.com/Flamefire/378f76254e62642ed760fd86a7771f7c for a full test report.

Flamefire · 2025-07-25T15:11:32Z

Ok, now I can't reproduce the failures anymore.

Crivella · 2025-07-25T15:54:28Z

Ok, now I can't reproduce the failures anymore.

The sanitizer tests can be flaky... considering you had 11/10 failures with 3 from the sanitizers i am not too surprised

Flamefire · 2025-07-25T16:07:34Z

So add "stack-overflow-with-asan.test" to the ignore list and call it a day?

The error from omp_host_pinned_memory is

"CUDA" error: Failure to alloc memory: Error in cuMemAlloc[Host|Managed]: invalid device context
"PluginInterface" error: Failure to allocate device memory: Failed to allocate from device allocator

and offloading/ctor_dtor.cpp just returns 2 without any output.

Crivella · 2025-07-25T22:43:05Z

Considering in the newer ECs we do skip_sanitizer_tests = True i would probably do it here as well.
I do not think the failing sanitizer tests will always be the same...

EG:

easybuild-easyconfigs/easybuild/easyconfigs/l/LLVM/LLVM-20.1.5-GCCcore-13.3.0.eb

Line 76 in 46bad6d

skip_sanitizer_tests = True

Crivella · 2025-07-29T11:54:45Z

@boegelbot please test @ jsc-zen3-a100
CORE_CNT=16
EB_ARGS="--installpath /tmp/$USER/pr23055"

boegelbot · 2025-07-29T12:00:11Z

@Crivella: Request for testing this PR well received on jsczen3l1.int.jsc-zen3.fz-juelich.de

PR test command 'if [[ develop != 'develop' ]]; then EB_BRANCH=develop ./easybuild_develop.sh 2> /dev/null 1>&2; EB_PREFIX=/home/boegelbot/easybuild/develop source init_env_easybuild_develop.sh; fi; EB_PR=23055 EB_ARGS="--installpath /tmp/$USER/pr23055" EB_CONTAINER= EB_REPO=easybuild-easyconfigs EB_BRANCH=develop /opt/software/slurm/bin/sbatch --job-name test_PR_23055 --ntasks="16" --partition=jsczen3g --gres=gpu:1 ~/boegelbot/eb_from_pr_upload_jsc-zen3.sh' executed!

exit code: 0
output:

Submitted batch job 7384

Test results coming soon (I hope)...

Details

- notification for comment with ID 3132130925 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

Flamefire · 2025-07-29T15:03:16Z

Test report by @Flamefire
FAILED
Build succeeded for 1 out of 2 (2 easyconfigs in total)
c113 - Linux AlmaLinux 9.4, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 560.35.03, Python 3.9.18
See https://gist.github.com/Flamefire/a4f958bcc1649ffb4e5f6b2d824ca051 for a full test report.

Flamefire · 2025-07-29T15:37:22Z

Test report by @Flamefire
SUCCESS
Build succeeded for 2 out of 2 (2 easyconfigs in total)
i7009 - Linux Rocky Linux 9.6, x86_64, AMD EPYC 7702 64-Core Processor (zen2), Python 3.9.21
See https://gist.github.com/Flamefire/1848dd59efb69ef085de7c1f0c7cfa75 for a full test report.

boegelbot · 2025-07-29T16:52:52Z

Test report by @boegelbot
FAILED
Build succeeded for 1 out of 2 (2 easyconfigs in total)
jsczen3g1.int.jsc-zen3.fz-juelich.de - Linux Rocky Linux 9.5, x86_64, AMD EPYC-Milan Processor (zen3), 1 x NVIDIA NVIDIA A100 80GB PCIe, 555.42.06, Python 3.9.21
See https://gist.github.com/boegelbot/1d93f56eedd9fc7067a96c388dc9e053 for a full test report.

Crivella · 2025-07-30T13:21:36Z

Umh we might also want to disable the fuzzer tests, when we set skip_sanitizer_tests

https://github.com/llvm/llvm-project/blob/36961202fbf45968cc273fa78fe3479409f5a9c7/compiler-rt/test/CMakeLists.txt#L66-L90

Right now through regex_subs = [(r'compiler_rt_test_runtime.*san.*', '')] we are only removing lsan ubsan sanitizer_common and the ones from the loop

Concerning the failing CUDA tests, not sure how relevant they are or if we want to add them to ignores tests in the EC ...

Flamefire · 2025-07-30T14:01:23Z

The "Failure to alloc memory: " sounds serious but they don't seem to be too consistent, so maybe not. I'm currently doing some rebuilds and will check if they always fail.

I also tried enabling the tests for the original EasyConfig (using Clang easyblock) but there it couldn't even finish compiling. Looks like a concurrency issue.

Crivella · 2025-07-31T08:31:56Z

@boegelbot please test @ jsc-zen3-a100
CORE_CNT=16
EB_ARGS="--installpath /tmp/$USER/pr23055"

boegelbot · 2025-07-31T08:40:12Z

@Crivella: Request for testing this PR well received on jsczen3l1.int.jsc-zen3.fz-juelich.de

PR test command 'if [[ develop != 'develop' ]]; then EB_BRANCH=develop ./easybuild_develop.sh 2> /dev/null 1>&2; EB_PREFIX=/home/boegelbot/easybuild/develop source init_env_easybuild_develop.sh; fi; EB_PR=23055 EB_ARGS="--installpath /tmp/$USER/pr23055" EB_CONTAINER= EB_REPO=easybuild-easyconfigs EB_BRANCH=develop /opt/software/slurm/bin/sbatch --job-name test_PR_23055 --ntasks="16" --partition=jsczen3g --gres=gpu:1 ~/boegelbot/eb_from_pr_upload_jsc-zen3.sh' executed!

exit code: 0
output:

Submitted batch job 7427

Test results coming soon (I hope)...

Details

- notification for comment with ID 3139036781 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

Flamefire · 2025-07-31T14:18:56Z

I think those failures aren't serious after all: Calling cuMemAlloc before any other CUDA stuff that creates a context is supposed to fail.
So I guess it is a bug in LLVM.

Comparing the EC built with Clang and LLVM and running the simple example:

With Clang the allocation succeeds but the free causes a ""CUDA" error: Failure to free memory: Error in cuMemFree[Host]: invalid argument"
With LLVM I get the same as in the test step ""CUDA" error: Failure to alloc memory: Error in cuMemAlloc[Host|Managed]: invalid device context"

So good to ignore I guess

Flamefire · 2025-07-31T20:32:07Z

Test report by @Flamefire
SUCCESS
Build succeeded for 2 out of 2 (2 easyconfigs in total)
c40 - Linux AlmaLinux 9.4, x86_64, AMD EPYC 9334 32-Core Processor (zen4), 4 x NVIDIA NVIDIA H100, 560.35.03, Python 3.9.18
See https://gist.github.com/Flamefire/099eec666d1403ea002be59606f5f449 for a full test report.

boegelbot · 2025-07-31T21:25:59Z

Test report by @boegelbot
SUCCESS
Build succeeded for 2 out of 2 (2 easyconfigs in total)
jsczen3g1.int.jsc-zen3.fz-juelich.de - Linux Rocky Linux 9.6, x86_64, AMD EPYC-Milan Processor (zen3), 1 x NVIDIA NVIDIA A100 80GB PCIe, 575.57.08, Python 3.9.21
See https://gist.github.com/boegelbot/eae48834eb468f50e6e1a09f16af9cdc for a full test report.

Crivella · 2025-08-04T08:22:02Z

@boegelbot please test @ jsc-zen3
CORE_CNT=16
EB_ARGS="--installpath /tmp/$USER/pr23055 Clang-18.1.8-GCCcore-13.3.0.eb python-igraph-0.11.9-foss-2024a.eb PySide6-6.7.2-GCCcore-13.3.0.eb pocl-6.0-GCC-13.3.0.eb"

Crivella · 2025-08-04T08:23:19Z

Doing one final test building all the software that currently exists in EB that depends on Clang-18.1.8-GCCcore-13.3.0

boegelbot · 2025-08-04T08:30:12Z

@Crivella: Request for testing this PR well received on jsczen3l1.int.jsc-zen3.fz-juelich.de

PR test command 'if [[ develop != 'develop' ]]; then EB_BRANCH=develop ./easybuild_develop.sh 2> /dev/null 1>&2; EB_PREFIX=/home/boegelbot/easybuild/develop source init_env_easybuild_develop.sh; fi; EB_PR=23055 EB_ARGS="--installpath /tmp/$USER/pr23055 Clang-18.1.8-GCCcore-13.3.0.eb python-igraph-0.11.9-foss-2024a.eb PySide6-6.7.2-GCCcore-13.3.0.eb pocl-6.0-GCC-13.3.0.eb" EB_CONTAINER= EB_REPO=easybuild-easyconfigs EB_BRANCH=develop /opt/software/slurm/bin/sbatch --job-name test_PR_23055 --ntasks="16" ~/boegelbot/eb_from_pr_upload_jsc-zen3.sh' executed!

exit code: 0
output:

Submitted batch job 7477

Test results coming soon (I hope)...

Details

- notification for comment with ID 3149606457 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

boegelbot · 2025-08-04T11:24:49Z

Test report by @boegelbot
SUCCESS
Build succeeded for 4 out of 4 (4 easyconfigs in total)
jsczen3c2.int.jsc-zen3.fz-juelich.de - Linux Rocky Linux 9.6, x86_64, AMD EPYC-Milan Processor (zen3), Python 3.9.21
See https://gist.github.com/boegelbot/d5036fea41bb090bdc95c7593cb14399 for a full test report.

Crivella

LGTM

passes both the CPU/GPU builds and also tested building software that depends on Clang-18.1.8-GCCcore-13.3.0 both as a builddep and normal dep

Crivella · 2025-08-04T11:40:15Z

Going in, thanks @Flamefire!

Use LLVM easyblock for Clang 18 easyconfigs

b490d00

github-actions bot added the change label Jun 6, 2025

This was referenced Jun 6, 2025

Deprecate the Clang Easyblock in favor of the LLVM one for newer (>=18.1.6) versions of Clang easybuilders/easybuild-easyblocks#3746

Merged

Move Clang easyconfigs to use LLVM easyblock #23028

Closed

Crivella mentioned this pull request Jun 16, 2025

Add easyconfig Clang-18.1.8-GCCcore-13.2.0.eb #22626

Closed

boegel added this to the 5.1.1 milestone Jul 1, 2025

boegel requested changes Jul 2, 2025

View reviewed changes

easybuild/easyconfigs/c/Clang/Clang-18.1.8-GCCcore-13.3.0.eb Outdated Show resolved Hide resolved

Flamefire and others added 2 commits July 2, 2025 15:13

Merge branch 'easybuilders:develop' into clang-llvm-18

6ed6a5d

Move easyblock param to top and add CUDA CCs

7a266f5

Flamefire mentioned this pull request Jul 3, 2025

Refactor LLVM easyblock and install OpenMP library symlinks easybuilders/easybuild-easyblocks#3815

Merged

boegel previously approved these changes Jul 3, 2025

View reviewed changes

boegel modified the milestones: 5.1.1, release after 5.1.1 Jul 4, 2025

Include OpenMP library aliases

d5a7eea

Flamefire dismissed boegel’s stale review via d5a7eea July 4, 2025 08:33

Skip failing tests

74161af

Skip sanitizer tests

92f1fc9

Ignore another test

a8831fe

Crivella approved these changes Aug 4, 2025

View reviewed changes

Crivella merged commit e04ee46 into easybuilders:develop Aug 4, 2025
8 checks passed

Flamefire deleted the clang-llvm-18 branch August 4, 2025 14:50

boegel mentioned this pull request Sep 23, 2025

Failing build of Clang-18.1.8-GCCcore-13.3.0-CUDA-12.6.0.eb #23973

Closed

Conversation

Flamefire commented Jun 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

boegel commented Jul 3, 2025

Uh oh!

boegelbot commented Jul 3, 2025

Uh oh!

boegel left a comment

Choose a reason for hiding this comment

Uh oh!

boegelbot commented Jul 4, 2025

Uh oh!

boegel commented Jul 4, 2025

Uh oh!

Flamefire commented Jul 4, 2025

Uh oh!

Crivella commented Jul 4, 2025

Uh oh!

Flamefire commented Jul 4, 2025

Uh oh!

Flamefire commented Jul 4, 2025

Uh oh!

Flamefire commented Jul 5, 2025

Uh oh!

Flamefire commented Jul 5, 2025

Uh oh!

Flamefire commented Jul 6, 2025

Uh oh!

Flamefire commented Jul 9, 2025

Uh oh!

Crivella commented Jul 9, 2025

Uh oh!

Crivella commented Jul 9, 2025

Uh oh!

Flamefire commented Jul 25, 2025

Uh oh!

Flamefire commented Jul 25, 2025

Uh oh!

Flamefire commented Jul 25, 2025

Uh oh!

Flamefire commented Jul 25, 2025

Uh oh!

Crivella commented Jul 25, 2025

Uh oh!

Flamefire commented Jul 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Crivella commented Jul 25, 2025

Uh oh!

Crivella commented Jul 29, 2025

Uh oh!

boegelbot commented Jul 29, 2025

Uh oh!

Flamefire commented Jul 29, 2025

Uh oh!

Flamefire commented Jul 29, 2025

Uh oh!

boegelbot commented Jul 29, 2025

Uh oh!

Crivella commented Jul 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Flamefire commented Jul 30, 2025

Uh oh!

Crivella commented Jul 31, 2025

Uh oh!

boegelbot commented Jul 31, 2025

Uh oh!

Flamefire commented Jul 31, 2025

Uh oh!

Flamefire commented Jul 31, 2025

Uh oh!

boegelbot commented Jul 31, 2025

Uh oh!

Crivella commented Aug 4, 2025

Uh oh!

Crivella commented Aug 4, 2025

Uh oh!

Flamefire commented Jun 6, 2025 •

edited

Loading

Flamefire commented Jul 25, 2025 •

edited

Loading

Crivella commented Jul 30, 2025 •

edited

Loading