Author: Robert Hyatt
Date: 20:50:55 06/19/01
Go up one level in this thread
On June 19, 2001 at 05:43:00, Gian-Carlo Pascutto wrote: >On June 18, 2001 at 21:07:19, Slater Wold wrote: > >>These tests were run on a Dual Pentium III 1,000mhz, 1028MB RAM. >> >>Doesn't seem like the MMX optimizations do anything. This was a tad >>surprising. > >They require a good superscalar MMX unit on the cpu to give an >advantage. The one on the Athlons flies on that code. I don't >know how the PIII does, but it wouldn't be a suprise if the Athlon >turns out to be more efficient. > >BTW. Your comparisation is horribly broken. If you bench you should >get the EXACT SAME nodecount for all versions. If they are searching >different trees, something is broken somewhere and there is little >sense in comparing the produced nps. > >-- >GCP I think he ran with mt=2.. that is _never_ a good way to benchmark, because the parallel search is highly non-deterministic and the node count can vary significantly, occasionally by a factor of more than 2x.
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.