Add configuration for AMD to EESSI-extend-easybuild.eb#206
Add configuration for AMD to EESSI-extend-easybuild.eb#206casparvl merged 9 commits intoEESSI:mainfrom
Conversation
cd8e08e to
7a0f948
Compare
ocaisa
left a comment
There was a problem hiding this comment.
LGTM but we will need to add some testing for this in https://github.com/EESSI/software-layer-scripts/blob/main/.github/workflows/tests_eessi_module.yml
4bc20c1 to
d1a699d
Compare
|
bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws for:arch=x86_64/amd/zen4 |
|
New job on instance
|
|
|
Wrong build command! bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws on:arch=zen4 for:arch=x86_64/amd/zen4,accel=amd/gfx90a |
|
New job on instance
|
|
|
bot: build repo:eessi.io-2025.06-software instance:eessi-bot-mc-aws on:arch=zen4 for:arch=x86_64/amd/zen4,accel=amd/gfx90a |
|
New job on instance
|
|
We will also need to add a check to the EasyBuild hooks for software that requires an AMD GPU: software-layer-scripts/eb_hooks.py Lines 714 to 738 in 0aaadf7 but this can be in a follow-up PR. Should be possible by taking the full list of dependencies (including toolchain deps which I think you can get from self.cfg._toolchain.tcdeps), and then just add adding ROCm-LLVM to acclerator_deps
|
|
For job 148716, the output is: Bit surprised about the easyconfig commit failure. But seems also the EESSI-extend also doesn't accept the accelerator target yet. |
|
Job has finished btw, bot just didn't report back. But because of the failure to load EESSI-extend, it starts to install with default EB configuration, meaning it's installing all de deps that are normally filtered |
|
Weird: I tried this locally, and I can't reproduce the issue. Here's what I tried: So that seems ok. I guess maybe it didn't rebuild EESSI-extend before building the easystack file? Or it did, but that wasn't used to build the easystack file? I'll double check the logs... |
|
Ah, I understand the problem (I think): it's essentially a catch-22. To install a new version of I'm honestly not sure why Alan's original build failed, because I think that's exactly what should be done: target a(/all) CPU targets. |
|
bot: build repo:eessi.io-2023.06-software instance:eessi-bot-mc-aws for:arch=x86_64/amd/zen2 |
|
New job on instance
|
|
bot:cancel jobid:150055 |
…b-5.3.0-001-system.yml
|
Deleted the rocm-based build, as we can't rebuild EESSI-extend AND target an accelerator at the same time. We'll just have to test this build locally - which I effectively did in #206 (comment) |
|
bot:cancel job:150055 |
|
bot: build repo:eessi.io-2023.06-software instance:eessi-bot-mc-aws for:arch=x86_64/amd/zen2 |
|
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
|
|
bot:status last_build |
|
This is the status of all the
|
6 similar comments
|
This is the status of all the
|
|
This is the status of all the
|
|
This is the status of all the
|
|
This is the status of all the
|
|
This is the status of all the
|
|
This is the status of all the
|
…er-scripts into eessi-extend-amd
|
I don't think we can implement the CI changes yet. Is probably because this new I.e. it is simply too early to implement this CI check. |
… for this to pass (e.g. the accelerator dir doesnt exist yet in CVMFS)
casparvl
left a comment
There was a problem hiding this comment.
Looks good to me, thanks @zerefwayne
CI will have to be updated later, after some other components have been fixed (the installation path exists for example)
|
Hmmm, I guess in theory we could keep this open, until we've done our first build (and the prefix exists). And then try to add the CI. At least we won't forget. I'll do that. The fact that this has been deployed already is no problem, CI isn't deployed anyway. |
We will have to be really careful in the and not merge other PRs that change EESSI-extend, otherwise your changes will be overwritten and things get messy... So maybe we should just merge this PR. |
|
Good point, I'll just open a new draft PR for updating the workflow |
|
CI will be updated in #214 |
No description provided.