[
https://issues.apache.org/jira/browse/ARROW-4547?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16769077#comment-16769077
]
Pearu Peterson commented on ARROW-4547:
---------------------------------------
[~Andrew_Palumbo], there are many possibly superfluous commands in your
`run_conda.sh`. I say "possibly" because it really depends what you need. For
instance:
Only one `conda create ...` command is required.
Only one `conda activate ...` command is required.
Is the aim to build pyarrow from source and just use it?
If yes, then you don't need `make format` as well as `-DCLANG_FORMAT_BIN=..` in
cmake. I am sure that then also some of the packages listed in `conda install
...` command are not required.
Also, then you would not need to use `python setup.py ... develop`, just
replace `develop` with `install` here.
Do you need numba.cuda support?
If not, then you don't need to define `NUMBAPRO_..` environment variables.
I think the most important parts for installing pyarrow with cuda support are:
Use `-DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX` when in conda environment.
Otherwise, use different prefix or default might be fine.
Usage of `-DARROW_PYTHON=on -DARROW_CUDA=on` in cmake and `python setup.py
build_ext --with-cuda ...`
Finally, you need to make conda command available to your gist.
> [Python][Documentation] Update python/development.rst with instructions for
> CUDA-enabled builds
> -----------------------------------------------------------------------------------------------
>
> Key: ARROW-4547
> URL: https://issues.apache.org/jira/browse/ARROW-4547
> Project: Apache Arrow
> Issue Type: Improvement
> Components: Documentation, Python
> Reporter: Wes McKinney
> Priority: Major
> Fix For: 0.13.0
>
>
> Building a CUDA-enabled install is not documented
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
