Olivier Teytaud: <[EMAIL PROTECTED]>: >> >> >> Yes. I use Sylvain's fpu and decrease it a little before starting a >> simulation, say, fpu *= 0.99. This is very simple and fast. > > >Ok. Perhaps I'm wrong (I might misunderstand your solution and I might be >wrong >whenever I've understood :-) ); but >- I think that this does not avoid redundancies in the tree, if I understand >well. >- also, even in a final node in a tree, it does not avoid redundancies for >moves > which have already been simulated. >- finally, in a final node in a tree and for moves which have not yet been >simulated, > it only avoid redundancies if 0.99 is sufficiently small.
Perhaps yes. I use _no_ pre-knowledge yet, ie, the initial value of fpu is a constant, 1.10. >Avoiding redundancies is very important in SMP parallelization. The limited >speed-up >of MPI parallelization in 9x9 is also probably due to redundancies, but for >that we have no >good solution - whereas in 19x19 (or even 13x13), the speed-up of the MPI >parallelization is great (both >in our results and other published papers), in 9x9 it is limited. Although I'm parallelizing in not SMP systems but a cluster of loosely coupled (small) computers connected through moderate speed networks using broadcasting positions, this may not change the vlaue of avoiding redundancies. I'll study more when implementing pre-knowledge or some. Thanks. Best regards, Hideki >Best regards, >Olivier >---- inline file >_______________________________________________ >computer-go mailing list >computer-go@computer-go.org >http://www.computer-go.org/mailman/listinfo/computer-go/ -- [EMAIL PROTECTED] (Kato) _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/