Translations:GPAW/13/en
Jump to navigation
Jump to search
This would use a single node with 8 MPI-ranks (ntasks) and 4 OpenMP threads per MPI rank (cpus-per-task) so a total of 32 CPUs. You probably want to adjust those numbers so that the product matches the number of cores of a whole node (i.e. 32 at Graham, 40 at Béluga and Niagara, 48 at Cedar or 64 at Narval).