pvbatch run on GPU

Hi,

I’m trying to run my pvbatch script on a GPU but with no success, I’m using:

pvbatch --egl-device-index=$CUDA_VISIBLE_DEVICES test.py

and I get:

e[0me[31m(   0.525s) [pvbatch         ]                       :0     FATL| Signal: SIGSEGVe[0m
error: exception occurred: Segmentation fault

Any idea on how to solve?

Regards

What does $CUDA_VISIBLE_DEVICES contains ?

It is pretty strange but the $CUDA_VISIBLE_DEVICES is not in my environment when I execute the code, the code is executed via slurm and nvidia module is loaded correctly as well as the GPU but the variable is not setted, is there a way to set it up by myself?

Proably, if you know the name of your egl devices.