{2025.06}[GCCcore/14.2.0] CPU-only dependencies for HIP and RCCL#1527
Open
zerefwayne wants to merge 7 commits into
Open
{2025.06}[GCCcore/14.2.0] CPU-only dependencies for HIP and RCCL#1527zerefwayne wants to merge 7 commits into
zerefwayne wants to merge 7 commits into
Conversation
casparvl
reviewed
Jun 18, 2026
| - fmt-11.2.0-GCCcore-14.2.0.eb | ||
| - gflags-2.2.2-GCCcore-14.2.0.eb | ||
| - glog-0.7.1-GCCcore-14.2.0.eb | ||
| - rocprofiler-register-0.4.0-GCCcore-14.2.0-ROCm-6.4.1.eb |
Collaborator
There was a problem hiding this comment.
I'm not sure if this should be here or not. I think it is cpu code, right? If so, then yes. But then why does it have ROCm as a dependency again...?
I'm also wondering if this will actually pass. I don't remember how the build scripts determine if something is a GPU build or not - I think it's based on whether CUDA/ROCm are dependencies, and if yes, it assumes it's a GPU build. Not 100% sure though. Feel free to give it a try...
Contributor
Author
There was a problem hiding this comment.
rocprofiler-register is based on ROCm-6.4.1, but we decided to build it using GCCcore, hence the versionsuffix. It doesn't have any dependency on ROCm-LLVM.
Collaborator
There was a problem hiding this comment.
I guess it's also a question of where we think this should be, and if the current setup in the build scripts make sense. I think for the nvidia stuff we decided to make a simple split: if it has CUDA as dep, then it's accelerator software. I guess we can (and should) do the same for ROCm. But let's first see if this fails today or not :)
Contributor
Author
There was a problem hiding this comment.
The build succeeded for Zen2, should we try building for all the archs? However, the deps with ROCm suffix don't need to be built for aarch64 😕
Collaborator
There was a problem hiding this comment.
Well, remove the zlib from the list first (see my other comment). Then, build for zen2 and one of the aarch64 targets. If that succeeds, build for the remaining targets.
While I agree that there's little point in having something that is only relevant in the ROCm context for aarch64 (rocprofiler-register and other ROCm-specific stuff), the rest (fmt, gflags, glob) is generic software that could be used outside of a ROCm context. And we should thus keep it identical.
If the aarch64 stuff fails on the ROCm-specific items in this easystack file, my suggestion would be to also split those off from this PR. Then, build this PR for all targets, and another one only for x86_64 targets.
boegel
reviewed
Jun 18, 2026
| - fmt-11.2.0-GCCcore-14.2.0.eb | ||
| - gflags-2.2.2-GCCcore-14.2.0.eb | ||
| - glog-0.7.1-GCCcore-14.2.0.eb | ||
| - rocprofiler-register-0.4.0-GCCcore-14.2.0-ROCm-6.4.1.eb |
Contributor
There was a problem hiding this comment.
@zerefwayne This one is not included yet in EasyBuild v5.3.0
Two options:
- use
from-committo pull it in from Add ROCm-6.4.1 easyconfigs easybuilders/easybuild-easyconfigs#25576 - wait until EasyBuild v5.3.1 is released and installed in EESSI (likely tomorrow)
I'm also not sure if this is really CPU-only...
It doesn't have any ROCM dependencies, but it is built from a rocm_* source tarbal...
@casparvl Thoughts?
Collaborator
There was a problem hiding this comment.
My thoughts are that I'm confused and don't know either XD I guess I need to find some time to have a closer look... :)
Collaborator
There was a problem hiding this comment.
I am slightly surprised that #1527 (comment) succeeded, with the whole rocprofiler-register having the ROCm suffix. But I guess since it does not depend on LLVM-ROCm, that's why it's considered CPU code - no matter the fact that this comes from the rocm sources.
The description at https://github.com/ROCm/rocprofiler-register suggests to me that this is CPU-only code: it's a helper library for intercepting API tables for profiling. Such API calls would happen on the host, so their intercepts would also happen on the host, I assume. The fact that it doesn't depend on LLVM-ROCm is a further proof that this is CPU only code as well. It's just... awkward that it's part of the rocm tarball. That makes me a bit uncertain whether this would work with any other ROCm versions. If it does, we could strip the suffix alltogether. If it doesn't, we should keep it.
At this point, I'm inclined to be conservative: if I don't know that it works with any ROCm version, I'm inclined to assume it only works with one - in which case I think the versionsuffix makes sense, as this easyconfig is tightly coupled to that ROCm version (6.4.1).
Contributor
Author
|
Test bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws for:arch=x86_64/amd/zen2 |
|
New job on instance
|
Contributor
Author
|
Test 2 bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws for:arch=x86_64/amd/zen2 |
|
New job on instance
|
Contributor
Author
|
Test 3 bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws for:arch=x86_64/amd/zen2 |
|
New job on instance
|
casparvl
reviewed
Jun 22, 2026
Co-authored-by: Caspar van Leeuwen <33718780+casparvl@users.noreply.github.com>
Collaborator
|
bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws for:arch=x86_64/amd/zen2 |
|
New job on instance
|
casparvl
requested changes
Jun 30, 2026
casparvl
left a comment
casparvl
left a comment
Collaborator
There was a problem hiding this comment.
Hmmm, since all of the from-commit's here are actually in the 5.3.1 release of EasyBuild... I think we should bump this list to use EB 5.3.1
Collaborator
|
bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws for:arch=x86_64/amd/zen2 |
|
New job on instance
|
Contributor
|
bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws for:arch=aarch64/neoverse_v1 |
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
boegel
changed the title
Contributor
|
JSC bot is down (system maintenance) |
Contributor
|
bot: build repo:eessi.io-2025.06-software instance:eessi-bot-jsc for:arch=aarch64/nvidia/grace |
|
New job on instance
|
Contributor
|
bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws for:arch=x86_64/generic |
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
|
New job on instance
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add CPU dependencies for HIP and RCCL #1526