[
https://issues.apache.org/jira/browse/ARROW-16464?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17619986#comment-17619986
]
Benson Muite edited comment on ARROW-16464 at 10/23/22 5:42 PM:
----------------------------------------------------------------
Probably easiest is to use GPUs that are hosted in the cloud.
[CircleCI|https://circleci.com/pricing/] does have an offering as does
[CIrun|https://cirun.io/]. Probably an estimate of the number of GPU minutes
needed per month would be helpful.
was (Author: baksmj):
Probably easiest is to use GPUs that are hosted in the cloud.
[CircleCI|https://circleci.com/pricing/] does have an offering as does
[CIrun[CIrun|https://cirun.io/]. PRobably an estimate of the number of GPU
minutes needed per month would be helpful.
> [C++][CI][GPU] Add CUDA CI
> --------------------------
>
> Key: ARROW-16464
> URL: https://issues.apache.org/jira/browse/ARROW-16464
> Project: Apache Arrow
> Issue Type: Bug
> Components: C++, Continuous Integration, GPU
> Reporter: Antoine Pitrou
> Assignee: Jacob Wujciak-Jens
> Priority: Critical
> Fix For: 11.0.0
>
>
> Arrow C++, PyArrow and perhaps other bindings have CUDA support, but none is
> currently tested on CI, and I think few of the contributors enable CUDA on
> their local builds.
> We should definitely exercise CUDA support, at least in the nightly builds
> where we may have more flexibility to use custom machines.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
