udacity/AIND-Isolation

Tournament code should be multithreaded

sunsided opened this issue · 2 comments

sunsided commented

When experimenting with new heuristics functions, most of the time is spent waiting on the tournament.py to finish, which takes ages even with a low amount of games (e.g. the default 5). Since it doesnt even appear to be CPU bound, evaluating matches in parallel would massively increase productivity.

cgearhart commented

Have a look at the experimental branch implementing multiprocessing for the tournament: https://github.com/udacity/AIND-Isolation/tree/xMultiProcessing

It hasn't been tested on Windows, and I won't have much time to support it for awhile (hence marking it "experimental"), but it should be most of the way there.

  • 👍1
sunsided commented

Very nice! Thanks!