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

conda-forge package for installation of non-Python dependencies? #494

Closed
darothen opened this issue Apr 27, 2021 · 5 comments
Closed

conda-forge package for installation of non-Python dependencies? #494

darothen opened this issue Apr 27, 2021 · 5 comments

Comments

@darothen
Copy link

Given this package's capability of running on the GPU via ThrustRTC, would it be possible to develop a conda-forge recipe to help users install all the pre-requisites for taking advantage of this hardware?

@slayoo
Copy link
Member

slayoo commented Apr 27, 2021

Thanks @darothen!

The only non-Python requirement should be the CUDA toolkit/driver. There seem to be a cuda-forge CUDA toolkit item: https://anaconda.org/conda-forge/cudatoolkit

@tehAgitto @Kyetuur @piotrbartman do you have any experience with Conda-forge?

@slayoo
Copy link
Member

slayoo commented Jun 20, 2021

@Golui has started exploring what needs to be done here.
First, it would be great to have ThrustRTC and CURandRTC available for conda, relevant requests reported:

@slayoo
Copy link
Member

slayoo commented Jun 23, 2021

Both ThrustRTC and CURandRTC are now available as conda packages! Thanks @fynv!

@slayoo
Copy link
Member

slayoo commented Sep 12, 2021

OK, I've tried to read through the [ana]conda resources, the issues linked from @Golui's PR, and made a few attempts to advance us here... but this only led me to one key question: what is the added value of providing the conda-forge package?

As pointed out by @Golui in the PR (#575), the conda meta files do not allow to provide information on which channel to get packages from. In our case, this is tricky as ThrustRTC and CURandRTC are available on the fyplus channel, chempy is available on the bjodah channel, while pyevtk is at the conda-forge channel. Numba, cudatoolkit, scipy and numpy are all available from different channels...

Furthermore, we do not really want to make cudatoolkit a dependency as PySDM is usable without a GPU using the Numba backend.

Finally, Anaconda users seem to be perfectly fine just opening the Anaconda prompt and typing pip install PySDM, right?

To be constructive here, I've just added to the PySDM README information for Anaconda users pointing out to the channels where the dependencies can be found, and to the conda docs section on using pip with conda: f649326

@darothen, would be great to hear your feedback (apparently we do not have any fans of conda among PySDM developers or users...). Thanks!

@slayoo
Copy link
Member

slayoo commented Sep 17, 2021

Let's close this and the PR for now. If there is any reason why conda users should not just open anaconda prompt and do pip install PySDM, please reopen

@slayoo slayoo closed this as completed Sep 17, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants