Author: Robert Hyatt
Date: 10:27:51 05/22/98
Go up one level in this thread
On May 22, 1998 at 03:34:18, Ulrich Tuerke wrote: >On May 22, 1998 at 00:53:36, Komputer Korner wrote: > >>On May 20, 1998 at 03:53:09, Ulrich Tuerke wrote: >> >>>On May 19, 1998 at 12:24:50, Komputer Korner wrote: >>> >>>>You are right. Comet is now winning the matches at 3 2 time controls. I >>>>now have a small opening book for Crafty and have put in the proper hash >>>>tables, so each side is using 24Mb main hash with Crafty also using 4Mb >>>>pawn hash. Actually I spoke too soon about the results of 60 moves/1 >>>>minute time controls. Comet actually had the upper hand in the end. >>>>However I am running these matches on a pentium 166 and Crafty does much >>>>better on Pentium Pros and above. Can you definitely verify that >>>>pondering is turned off for Comet during these matches in Winboard? I >>>>have pondering turned off for Crafty but I don't know how to turn the >>>>pondering off or whether it is by default off for Comet. If Comet is >>>>pondering then these matches aren't fair. >>> >>>Unfortunately, I am not really familar with WINBOARD. By default, >>>Comet's pondering should be turned on. ("Pondering" means use of >>>opponent's time too, right ?) >>>If you can manage to send an "easy" command via the WINBOARD interface >>>to Comet, then it will turn off pondering. >>> >>>But I think it would also be fair to let both programs ponderm, wouldn't >>>it ? >> >> >>When both programs ponder I get error messages and vanishing pieces when >>running in WIN NT 4 SP 3. Also even when only Comet is pondering, The >>first Comet process doesn't exit out of RAM when the next game starts. >>Thus half way through a match, there are many Comet processes in RAM >>each taking up CPU time. How would you set pondering off on a command >>line with Winboard? There are other problems with these matches such as >>both programs letting their clock run out even when in Fischer time >>controls. > >Peculiar! I have made 40-50 games between Comet and TCB, GNU, Crafty, >and Patzer on my PC at home (sudden death mode) - always both programs >pondering. No problem. I've used WIN-95. Roland Pfister didn't report >any problem either on his WIN-NT system. >Don't know what's going on. May be, winboard (or is it windows ?) is a >rather complicated system ? There is more to this than meets the eye I'm afraid. In "stress tests" I play game/1sec games between two crafty processes, just to bang on everything as hard as possible. At this rate, I average about 90 games per minute, and I have let it run for hours with no hangups. And then someone reports that when *they* run crafty vs crafty, strange things happen. Ditto for Crafty vs TCB or whatever... Therefore, I'm not sure who/what's at fault. I'm convinced something is wrong, but since I can't reproduce it with xboard, I have no chance of fixing it with no clue as to what is happening, *if* it is a bug in my code. But I can't imagine a harder test than game/1sec, two programs, both pondering, etc. That is a real test and it is fun to watch the board simply "blur" and then reset, then 'blur" and then reset. I set this up by using the crafty.rc file and setting the search time to .01 seconds... It can be nearly done by -tc 1 -mps 999 which says (roughly) 100 moves per 10 seconds... But we need winboard debug output for a game where a piece disappears or where the illegal move popups show up... then we can see who is producing the illegal move or why winboard/xboard thinks a move is illegal...
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.