Number of CPU threads

[Engineering-Fluid-Dynamics]

To whom it may concern,
Currently, for CPU-based calculations, the number of threads must be manually set in each terminal using export OMP_NUM_THREADS. It would be very convenient if this could instead be configured as a parameter in the settingsDict, or even better, directly within couplingProperties for coupled calculations. Ideally, it should allow specifying whether the simulation is parallel and, if so, define the number of cores (threads) to be used. Thanks.

[PhasicFlow]

This is a good option to be added. I suggest the following sub-dict to be added to couplingProperites file.

// location: in coupling properties 
cpuLoad
{
	type maximum;
	threadMPIJob 2;
}

maximum: uses maximum load on the CPU (equal to maximum number of threads)
adaptive: based on the number of particles and number of MPI jobs
manual: user specifies number of threads per each MPI job 

We have three execution modes in DEM side: Serial, GPU, and OpenMP

1. Serial: all the above options have no effect, since the DEM side is working in serial mode so does the coupling part.
2. GPU:

• maximum: sets number of threads to max_threads
• adaptive: sets number of threads to max_threads/MPIJob
• manual: by user

3. OpenMP:
   • maximum: no changes to the number of threads, since DEM side is probably using maximum number of threads
   • adaptive: sets number of threads to max_threads/MPIJob and set it back to the original value when coupling calculations have been finished (number of particles is also effective).
   • manual: sets the value specified by user, and set it back to the original value when coupling calculations have been finished.

any comments?