cluster computing - code flow of slurm gpu allocation -
does 1 know code flow how gpus allocated in slurm? have gone through , found cuda_visible_devices environment variable not updated in code. how done in code?
my goal add new hardware support using environment variable allocation.
take @ gpu gres plugin. there code modifies cuda_visible_devices environment variable. file path is:
src/plugins/gres/gpu/gres_gpu.c