Embarrassingly parallel processing with OCC

[cvoivret]

Hi, I have developed a python script to compute shadow on IFC model thanks to OCC. This script apply the same processing (CAD operations) to different faces of the model resulting on the creation of a new face without topologic link with the model. I apply this process numerous times with sligthly different parameters. I have tried to parallelize the processing (batch) with a Pool or with joblib without success. I suspect that because of the memory management model of OCC in conjunction of python wrapping, these simple approaches of parallel processing are doomed. Does anybody have any tough on a possible way to success (preferably in python ;-) ) ?