On Sun, Jun 13, 2010 at 5:31 PM, Mark Boon <[email protected]> wrote: > So far all numbers seem to confirm my initial statement that CPUs > didn't get all that much faster during recent years. The focus is more > on getting more cores working together faster. > > I think a lot could be gained if multi-threading could be used on a > more fine-grained level. For that we need faster task-switching. If > task-switching could be done in a microsecond instead of hundreds of > microseconds then it would be much easier to take advantage of larger > numbers of cores. > > But there are more bottlenecks to overcome to make massive parallelism > a reality of course. > > Mark
I thought one of the best features of MC methods is that they parallelize easily. MCTS does you some information from previous playouts to bias the current one, but I can't imagine that the performance would suffer much by having this information slightly outdated. Does anyone have a quantitative comparison of a single-threaded MCTS against a multi-threaded version of the same program, both searching the same number of playouts? Álvaro. _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
