Mark Boon: <[email protected]>: >On Tue, May 24, 2011 at 10:30 AM, Michael Williams ><[email protected]> wrote: >> Yeah, that seems like it would work well. And you could probably do many >> playouts at each leaf, instead of one. >> > >From the little I know about GPU programming, I understood that in >order to get good parallelism you'll have no choice but to do many >simulations simultaneously to make it worth it. > >But for MCTS it's not very efficient to simulate the same leaf many >times at once. So it will probably only be worth it if you can get in >a fairly large number of simulations at similar cost to one simulation >done by a CPU.
That's not correct. Leaf parallelism is efficient enough and can compete root parallelism. Past bad results are mainly due to synchronizing all simulations invoked from a leaf. Hideki >Mark >_______________________________________________ >Computer-go mailing list >[email protected] >http://dvandva.org/cgi-bin/mailman/listinfo/computer-go -- Hideki Kato <mailto:[email protected]> _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
