Olivier Teytaud: <[EMAIL PROTECTED]>: >> >> >> 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. > > > >Do you have the winning rate of your best parallel algorithm against the >sequential version in 9x9 Go >in such a setting without shared memory ? We stay below 80% for this form of >parallelization (but this >speed-up can be cumulated with SMP speed-up), whenever we use fast networks >(infiniband).
Not yet. I have many benchmarks now but only against GNU Go level _0_ on 9x9 and 13x13. The best WR on 9x9 is 72% using four quad-core PCs on a GbE private LAN with 0.16 second per move (less than 10 seconds per game) but the speed-up (strength-speedup by G. Chaslot) against 1 PC is just 1.6. This may be improved in longer time settings but I'm not sure. # I can use only 4 x 4 cores in hand now so that it's very hard to have benchmarks on 19x19. >In 19x19, it's much better, but the MPI parallelization of 9x9 Go is >challenging. Strongly agree. Even in 13x13, it's much better. Best, 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/
