Author: Robert Hyatt
Date: 06:10:48 05/27/98
Go up one level in this thread
On May 27, 1998 at 03:34:03, Moritz Berger wrote: >On May 27, 1998 at 01:26:19, Mark Young wrote: > >>On May 26, 1998 at 23:53:22, Robert Hyatt wrote: > >>>would be interesting to see who/how/what they've done with the parallel >>>search, since none of 'em have been "practicing" parallel search. I >>>have >>>an idea. :) >> >>Please explain. any insight would be interesting. > >Well, obviously Fritz 5 has a k-variant mode. You could just take that >and drop it on multiple CPUs - not really parallel search, but speeds up >things if the 2nd best move turns out to be better ... (different >speculations welcome) > >Moritz doesn't work well at all. IE in Crafty, the first move typically takes 50-75% of the total search time. Searching two root moves in parallel won't even speed things up by 50% n that case. SMP is non-trivial, and if the engine is assembly, it is *really* non- trivial (as we found in Cray Blitz, for example). But in any case, searching root moves in parallel will not give much of a speedup at all, if it gives any.
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.