Author: Bruce Moreland
Date: 11:37:18 10/13/98
Go up one level in this thread
On October 12, 1998 at 14:25:08, Ernst A. Heinz wrote: >With the help of its interior-node recognizers and its endgame databases >"DarkThought" solves 23/24 at 60 sec. per move on a 600MHz Alpha-21164a. I would like to ask you to supply more details about the recognizers (yes, I read your paper). When you use tables at the tips, you need to have some means of dispatching into the right table. If the tables aren't present, you can use the same dispatch mechanism to get you into the right routine appropriate to the situation. So, for instance, mine understands that it is supposed to use a table for KRN vs KR, but if the table isn't present, it goes to some other routine that returns something close to a draw value. I invented this myself, but I am sure that others invented it independently. Do you have special routines for higher material cases, for instance K+P endings with many pawns, or rook and pawn endings with many pawns? bruce
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.