Parallelizing your calculations

For precise simulations on big grids, the computation can take a few minutes, especially if designs with counterpropagating beams are used. In order to speed up the optimisation process, we implemented a native way for dispatching the computation on various cores of the processor. This is done thanks to the ParallelSimulator class, which takes advantage of the Python multiprocessing package.

After setting the simulation and before running Simul.compute, you can set up a ParallelSimulator that will handle the parallelization of the computation:

   Simul = Simulation(syst, SiO2(), trap, datafolder,surface)
   Simul.simulator = ParallelSimulator(max_workers=None)

max_workers specify the number of maximum cores that will be used for the simulation. When set to None, all the possible cores will be used. By default an object SequentialSimulator is created, avoiding any issue that could arise when using multiprocessing.

On Linux, the htop command allows to see the usage of the microprocessor cores.