Author: Robert Hyatt
Date: 20:44:05 11/30/02
Go up one level in this thread
On November 29, 2002 at 22:20:49, Bob Durrett wrote: >On November 28, 2002 at 04:07:55, Aaron Gordon wrote: > >>http://speedycpu.dyndns.org/crafty/craftybench4.jpg > > >Presumably, the chart referenced above used "an average position." Maybe the >Kn/s was measured for a suite of test positions and then an average calculated? > The Crafty "benchmark" command does this. It runs six different positions, covering opening middlegame and endgame positions. The result is the average KNPS for the six positions. POsitions can make a huge difference in NPS. IE crafty is slowest early in the game where it is evaluating development and castling issues that are slow to compute. Other problems arise where things are wild and large eval terms are generated, reducing the effectiveness of lazy evaluation... And then there are positions with lots of deep checking lines that kill NPS... >Just curious as to how much difference Kn/s varies due to the different types of >positions encountered? > >Are there extremes and special positions for which Kn/s is very different from >usual? > >Bob D.
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.