though in the long term i should convert these workflows to be agnostic to the system and scheduler, for now it is convenient to have them pre-written for the USF hpc. My former default method of parallel cpus is not ideal, because --ntasks can spread the number of assigned cpus across multiple nodes, and almost all of my scripts use multithreading, not mpi. i should intead specify that i only want a single node per script, and use --cpus-per-node for multithreading purposes.
though in the long term i should convert these workflows to be agnostic to the system and scheduler, for now it is convenient to have them pre-written for the USF hpc. My former default method of parallel cpus is not ideal, because --ntasks can spread the number of assigned cpus across multiple nodes, and almost all of my scripts use multithreading, not mpi. i should intead specify that i only want a single node per script, and use --cpus-per-node for multithreading purposes.