In a FENSAP/DROP3D sweep run, or a DROP3D Langmuir run, if the number of CPUs per run is lower than the total number of CPU allocated for the execution, multiple instances of FENSAP/DROP3D will run in parallel.

Example 21.1: MPI Settings for 48 CPUs

Example 21.2: Sweep Solver Settings

This example shows that 6 parallel execution of FENSAP will be executed (would require 6 seats of FENSAP-ICE).

FENSAP-TURBO will also execute multiple parallel jobs. The CPU subdivision among the rows of the turbomachine is configured in the FENSAP-TURBO Run panel. The number of cores used for each row depends of the machinefile used.

Multi-core MPI Machine:

If the job is executed on a single machine, with no queuing system nor a machinefile, all the FENSAP processes will be launched on the same machine. For optimal execution speed, the machine should have enough CPU-cores to suit the job.

Hyperthreading:

A machine with 8 physical cores and enabled hyperthreading will display as 16 cores in the task manager, however running with more than 8 CPU-cores on this machine will typically not lead to faster execution speed.

Machinefile:

If the MPI settings for the job contain the -machinefile parameter, the specified file will be subdivided to suit the multiple parallel jobs. Some queuing system won’t require the -machinefile argument in the MPI settings; FENSAP-ICE will then automatically work on $PBS_NODEFILE (PBS) or $PE_HOSTFILE (SGE).

Job #1: host1 (4 CPUs)

Job #2: host2 (2 CPUs) + host3 (2 CPUs)