submit_benchmarks_legi.py

from fluiddyn.clusters.legi import Calcul7 as Cluster
# from fluiddyn.clusters.legi import Calcul8 as Cluster

cluster = Cluster()
cluster.commands_setting_env.extend([
    'module load p3dfft/2.7.4-mt',
    'module load pfft/1.0.6'])


def submit(nb_nodes, nb_cores_per_node=None):
    if nb_cores_per_node is None:
        nb_cores_per_node = cluster.nb_cores_per_node
    nb_mpi = nb_cores_per_node*nb_nodes
    cluster.submit_command(
        'fluidfft-bench 1024 -d 2 '
        '-o /.fsnet/data/legi/calcul9/home/augier3pi/fluidfft_bench',
        name_run='fluidfft-bench_{}'.format(nb_mpi),
        nb_nodes=nb_nodes,
        # nb_cores_per_node=nb_cores_per_node,
        nb_cores_per_node=cluster.nb_cores_per_node,
        walltime='00:04:00',
        nb_mpi_processes=nb_mpi, omp_num_threads=1,
        ask=False,
        delay_signal_walltime=None)

nb_nodes = 1
for nb_cores_per_node in [2, 4, 8, 12, 16, 20]:
    if nb_cores_per_node > cluster.nb_cores_per_node:
        continue
    submit(nb_nodes, nb_cores_per_node)

for nb_nodes in []:
    submit(nb_nodes)