Author: Robert Hyatt
Date: 18:04:58 05/05/01
Go up one level in this thread
On May 05, 2001 at 16:54:03, Jesper Antonsson wrote: >On May 05, 2001 at 10:51:00, Robert Hyatt wrote: >>I missed that last question. For crafty, "raw NPS" is simply N times faster >>than one processor. The .7 factors out the roughly 30% extra work each >>processor does in a parallel search. IE it would be pretty equivalent to the >>numbers Hsu reports when he says 200M. > >How come he got a .2 factor and you get .7? I can understand that a quad machine >lose less than their 32/512 processor (or was it 256?) monster, but on that 64 >processor machine you're going to play with you are getting close to their kind >of massive parallelism. > >Jesper Different search methods. He has a two-level parallel search, the first N plies are done on one cpu, the next N are done on the 32 SP nodes in parallel, the last N are done in hardware processors (again in parallel). It is harder to make that kind of two-level search as efficient as a pure shared-memory SMP search...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.