Reenable unit tests compilation and execution on all CI platforms. about geos HOT 7 OPEN

Can you give me any insight into the dev workflow to put together the CI docker environments? I know AMD provides HIP/ROCM containers so it should be doable without too much difficulty beyond building our TPLs I would think.

If it's anything like nvidia, they have pre-built images with CUDA installed already:
https://hub.docker.com/r/nvidia/cuda

Then it's a matter of building our dependencies on top of the pre-built images:
https://github.com/GEOS-DEV/thirdPartyLibs/blob/master/docker/clang-cuda/Dockerfile

from geos.

The documentation suggests no:

https://docs.github.com/en/actions/using-github-hosted-runners/about-github-hosted-runners#supported-runners-and-hardware-resources

from geos.

I can put together a PR for this, as I also want to get started on adding a CI that covers the HIP/ROCM build.

@bmhan12 Can you give me any insight into the dev workflow to put together the CI docker environments? I know AMD provides HIP/ROCM containers so it should be doable without too much difficulty beyond building our TPLs I would think.

from geos.

Do those GH Actions runners actually come with a usable GPU? Some of our unit tests will launch device kernels I believe.

from geos.

Did our travis platforms have physical GPUs?

from geos.

Did our travis platforms have physical GPUs?

No, they didn't. I think that's why we were only building the tests and not executing them. I guess we can just re-enable the building of the tests for now and add the execution once we have the local machine ready.

from geos.

@TotoGaz was this resolved after moving into GHA?

from geos.