Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

document CUDA builds #1226

Merged
merged 8 commits into from
Jan 8, 2021
Merged

document CUDA builds #1226

merged 8 commits into from
Jan 8, 2021

Conversation

jaimergp
Copy link
Member

Will close #901

@jaimergp jaimergp requested a review from a team as a code owner December 20, 2020 20:44
@beckermr
Copy link
Member

@jakirkham @kkraus14 can you double-check the CUDA info in this PR?

Copy link
Member

@ocefpaf ocefpaf left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@jaimergp I'm approving mostly the text and readability b/c I'm not too familiar with the process used here.

@jakirkham or @kkraus14 can you take a look and merge?

Copy link
Contributor

@kkraus14 kkraus14 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, one suggestion

src/maintainer/knowledge_base.rst Outdated Show resolved Hide resolved
Copy link
Member

@jakirkham jakirkham left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good. Thanks Jaime! 😄

Made a few minor suggestions above. Also generally agree with the points others have made here.

jaimergp and others added 2 commits December 22, 2020 12:43
Co-authored-by: jakirkham <jakirkham@gmail.com>
@jaimergp
Copy link
Member Author

jaimergp commented Jan 5, 2021

Just noting that this is not ready to merge yet, since some of the workarounds here described wouldn't apply after conda-forge/nvcc-feedstock#58 is approved and merged.

Copy link
Member

@jakirkham jakirkham left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks again for the work here Jaime! 😄

Included a couple minor suggestions below.

src/maintainer/knowledge_base.rst Outdated Show resolved Hide resolved
src/maintainer/knowledge_base.rst Outdated Show resolved Hide resolved
jaimergp and others added 2 commits January 8, 2021 10:12
Co-authored-by: jakirkham <jakirkham@gmail.com>
Comment on lines +1221 to +1224
build:
...
missing_dso_whitelist:
- "*/nvcuda.dll" # [win]
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there any way to add this the global conda-forge settings?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not that I know of.

@jakirkham jakirkham requested review from isuruf and beckermr January 8, 2021 18:46
@jakirkham
Copy link
Member

Would be good if people can take another look 🙂

Copy link
Member

@beckermr beckermr left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sometime in the future it would be good to document what we now think about gcc versions and cuda versions.

cc @jakirkham @kkraus14 @leofang @isuruf

Copy link
Member

@leofang leofang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Great work here and elsewhere in the CF territory, @jaimergp! LGTM. I left a question for everyone here, but it need not be addressed in this PR.

Comment on lines +1151 to +1154
* ``nvcc``: Nvidia's EULA does not allow the redistribution of compilers and drivers. Instead, we
provide a wrapper package that locates the CUDA installation in the system. The main role of this
package is to set some environment variables (``CUDA_HOME``, ``CUDA_PATH``, ``CFLAGS`` and others),
as well as wrapping the real ``nvcc`` executable to set some extra command line arguments.
Copy link
Member

@leofang leofang Jan 8, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I understand this PR is for recipe maintainers, but I am wondering if we could mention this somewhere that the nvcc package is not for end users to install and use (as it's just a wrapper over the real compiler, which is missing)? I've seen several bug reports (including my own) in which the users do not have a local CUDA Toolkit installation and mistakenly think installing nvcc would enable them to compile.

@leofang
Copy link
Member

leofang commented Jan 8, 2021

Sometime in the future it would be good to document what we now think about gcc versions and cuda versions.

@beckermr What specifically do you have in mind?

@h-vetinari
Copy link
Member

Sometime in the future it would be good to document what we now think about gcc versions and cuda versions.

@beckermr What specifically do you have in mind?

Likely the interactions/restrictions being discussed in conda-forge/nvcc-feedstock#51, conda-forge/conda-forge-pinning-feedstock#1000, etc...?

@beckermr
Copy link
Member

beckermr commented Jan 8, 2021

Yup @h-vetinari has the right idea. There has been back and forth on this. As more people start to use CUDA builds, there will be more questions etc.

@leofang
Copy link
Member

leofang commented Jan 8, 2021

I see. Indeed it makes sense. Thanks @h-vetinari @beckermr. We should probably address it (including conda-forge/conda-forge-pinning-feedstock#1053) before the CUDA 11.1 migration starts.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Development

Successfully merging this pull request may close these issues.

Documenting GPU builds
9 participants