[Enhancement] GIL release about scikit-fmm HOT 2 CLOSED

OK, good catch, that would have been hard to track down later. I think option 3 is out, this is quite dangerous. Like the idea of having an optional argument allow_threading which makes copies of the input arrays and releases the GIL inside the C++ code (toggles your options 1 and 2). It should default to False to keep the existing behavior for existing code. It would be easy to create the copies with np.copy in the pre_process_args in pfmm.py. I would think doing the copy in Python is still fast and it keeps the C++ code simple.

from scikit-fmm.

I thought about this a little more and decided not to pursue this. Shared memory multi-threading is difficult to do correctly and would add complexity to the library. At least for now, the multiprocessing library should allow people to make use of multi-core computers for parametric studies.

from scikit-fmm.