Problem with multiprocessing.

[joakim-hove]

Problem with multiprocessing.

[joakim-hove]

The problem is that the multiprocessing is based on process-based parallellism, and the Python objects are passed from child process to parent through Python pickle functionality - that does not (and can not) carry over the pointers to the underlying C structures - that is way it crashes.

To achieve your goal you would have to use thread based parallelism - slightly more complex, but perfectly doable. Might be able to assist you with later, but not right now.

[wouterjdb]

Ok, I'm going to give this a try.

[wouterjdb]

Instead of

from multiprocessing import Pool

I used

from multiprocessing.pool import ThreadPool as Pool

in combination with the code I supplied in the original Issue.

This works and gives some speed improvements. Thanks for the tip.

[pgdr]

I'm closing this as I can't see any concrete issue to fix.

Please reopen and rewrite the issue if there still is one.