WIP: Add Linux CUDA builds #9804
Conversation
[skip lint]
[skip lint]
|
We still need to update this to identify and use the right variant. |
|
Subscribed! Let me know how can I help. |
| strategy: | ||
| maxParallel: 1 | ||
| matrix: | ||
| linux_cuda-9_2: |
There was a problem hiding this comment.
@jakirkham, I don't understand. This is adding 3 more azure jobs, right?
There was a problem hiding this comment.
It adds 1 job that sequentially builds each supported CUDA version.
There was a problem hiding this comment.
Ah, I guess jobs name in azure is slightly different from what is in travis.
Is there a way to skip these builds if there is no cuda needed?
There was a problem hiding this comment.
You can do a fan out as in conda-forge/_azure_pipelines_test#3
You can do the first job which is the existing one and set the variable NEED_CUDA to 0 or 1 depending on whether cuda compiler is used or not.
Then the 3 cuda jobs can run in parallel if NEED_CUDA is 1 and if not they are skipped.
There was a problem hiding this comment.
Yeah I was wondering that same question, but didn't have a good answer at the time. Thanks for investigating.
Fanning out seems reasonable. Guessing this will require an initial parsing of the recipes collected here before building, do you know if there is a good way to hook into the parsing process here to get that information?
There was a problem hiding this comment.
You can, but if you skip cuda_compiler_version == None, then it wouldn't work. So do the easiest way and check for - {{ compiler('cuda') }} in the recipe meta.yaml. There'll be false positives (if that line is commented), but there won't be false negatives. This can be done in the build_all.py where you can just echo that magic line.
|
Hi friend! We really, really, really appreciate that you have taken the time to make a PR on In an effort to maintain this repository and increase the signal-to-noise for open PRs, the maintainers of If you'd like to keep it open, please comment/push and we will be happy to oblige! Note that very old PRs will likely need to be rebased on master so that they can be rebuilt with the most recent CI scripts. If you have any trouble, or we missed reviewing this PR in the first place (sorry!), feel free to ping the team using a special command in a comment on the PR to get the attention of the Cheers and thank you for contributing to this community effort! |
|
Hi friend! We really, really, really appreciate that you have taken the time to make a PR on In an effort to maintain this repository and increase the signal-to-noise for open PRs, the maintainers of If you'd like to keep it open, please comment/push and we will be happy to oblige! Note that very old PRs will likely need to be rebased on master so that they can be rebuilt with the most recent CI scripts. If you have any trouble, or we missed reviewing this PR in the first place (sorry!), feel free to ping the team using a special command in a comment on the PR to get the attention of the Cheers and thank you for contributing to this community effort! |
|
Done in #13469 |
Fixes #9773
Merges #9802
This adds Linux CUDA builds for the supported versions of CUDA. These are part of a matrix job that builds each CUDA version sequentially to avoid using too many jobs when these are not needed. Uses the same build images and overall strategy that is used in the feedstocks.
Checklist
url) rather than a repo (e.g.git_url) is used in your recipe (see here for more details)