Author: leonid
Date: 11:55:40 12/03/00
Go up one level in this thread
On December 03, 2000 at 13:25:00, Paul Doire wrote: >I am attempting same pc engine matches. It has been brought to my attention >that Didziz has a page that describes the procedure for proper comparison. >I have attempted to implement his suggestions(comparing think times with other >prog running in background). >But CM6000 doesn't have ability in "think lines" to measure nodes, only depth >such as 4/09,5/10.Rebel 11 and CM8000 do have the ability to measure nodes. >Is there a universal nodes measurement utility available so I can properly >gauge nodes per second??? > >I have attempted "Depth" searches but am unsure if accurate,Chess Tiger >searches to depth of ten with CM6000 running in background in only 2.6 secs. >While CM6000 takes 42 secs to get to 5/10 with Chess Tiger in background. >Is it possible for Chess Tiger to search over 16 times faster??? > >Thanks, >Paul Can't respond nothing to your question but would like to ask you something. Please indicate position that you tried and exact time for each program that solved it. Say nodes/per/second, computer speed and which way search what done (brute force, selective, hash table...). Even if I am not now in programming (have other things to do for a while) I am still curious about NPS indicator. In my program this counting is done in very simple way. I say this just for the case that it could be useful. For count goes each position (node) that is used for recognition of real value for given ply. If in certain ply 30 moves (nodes) were created (almost all the time legal moves) but only 6 were used to find value for this ply, only 6 will be counted. In my program, when search is done by brute force in the middle of the game, around 5 to 6 moves are created for each one that will be used and counted. Program don't use hash table. Leonid.
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.