Author: Ernst A. Heinz
Date: 06:57:19 05/11/98
Go up one level in this thread
On May 11, 1998 at 09:40:18, Robert Hyatt wrote: > [...] > >I've seen no results that don't get somewhat worse with more processors, >excepting for Feldman's "young brothers wait" which starts off fairly >badly >but doesn't get worse. (by badly I mean that with 2 processors, It may >only >be 1.5-1.7 times faster, where Cray Blitz was almost dead on 2x >faster...) My personal impression is that Feldman's *relative* speedup results look good because he used one of the worst sequential programs imaginable (with much work to do you always get nice *relative* speedups -- just trace the fixed-depth node counts of the BK test set in his Ph.D. thesis) and because he did not factor the increasing sizes of his distributed transposition tables into the final calculation ... The really interesting numbers, namely *application* speedup of "Zugzwang" as compared with a well-tuned sequential program, ought to be horrible. =Ernst=
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.