SLURM gres option

When one tries to run unit tests on GPU or run MMS on GPU using submit_intel_gpu.sh or submit_gnu_gpu.sh, one must pay attention to the SLURM generic resource option --gres and set it accordingly the GPU architecture on the machine. Always take a look at the documentations of the machine for correct SLURM options.

In MPCDF machines: --gres=gpu:<GPU_ARCH>:<NUM_GPU>,

• <GPU_ARCH> : GPU architecture, i.e. in cobra v100, in raven a100

• <NUM_GPU> : Number of GPU to reserve, i.e. max number in cobra is 2 and in raven is 4