Support for multiple GPUs on local executor

[petroslk]

Hello!

Is there a plan to integrate GPU scheduling into the local executor? We need to use nextflow on a large workstation with multiple GPUs. There was a solution proposed at #997 but I am not sure this is on your radat at the moment. I guess something along the lines of using CUDA_VISIBLE_DEVICES and nvidia-smi to track what each GPU is running would do the trick? Is there any workaround available at the moment?

Thanks

[andreas-wilm]

For reference see also discussions here: https://community.seqera.io/t/proper-way-to-utilize-1-gpus-in-a-single-machine/455/7