Hi Tian,
Generally initialization should not take long compared to the simulation time (especially initializing source particles which is a relatively simple operation). It is true that most of the work during initialization is done in serial, so it not surprising that you only see one thread being used. This is usually not an issue because the overall initialization time should be small. Would you be willing to share your model so that I can try to see what might be going on?
Best regards,
Paul