Author: Robert Hyatt
Date: 21:19:00 10/22/04
Go up one level in this thread
On October 21, 2004 at 23:47:40, Stuart Cracraft wrote: >On October 21, 2004 at 22:49:28, Mathieu Pagé wrote: > >>I just read this paper by Robert Hyatt on rotated bitmap: >> >>http://www.cis.uab.edu/info/faculty/hyatt/bitmaps.html >> >>It describe how to generate attack_to[] and attack_from[] bitmaps. I think those >>tables are not updated incrementaly and are computed from scratch at each node. >>Am I right, or is there a efficient way to do it incrementaly ? >> >>Moreover, it seem to me that attack_to should be computed "on demand" only, >>since we will probably not use every square of it. So it could be a function : >>attack_to(int sq). Does this make sense ? >> >>Mathieu Pagé >>mathieu.page at gmail dot com > >I don't know anything about Bob's URL but incremental updates should >only require work on some of the pieces. > >Bob has made homage to Slate/Atkin of Chess 3.x/4.x at inception of his Crafty >effort as I recall from another writing. > >I sure hope it is incrementally updated since having taken only the >64-bit part of Slate/Atkin wouldn't be the whole enchilada. If he decided >not to do incremental updates, I'd be curious what he calculated the cost >factor of doing so vs. not doing so to be. The non-incremental version was 4x faster. That was what led me to the concept of "rotated bitmaps". The incremental cost was way too high. version 6.0 of crafty threw out the incremental stuff and was the first rotated bitmap program around... > >A quick search of Crafty's main.c reveals minimal mention >of the word "incremental"... > >Too bad and I liked Chess 0.5 in Byte too. Never got it working though. >Regrettable that those giant monster's Chess 4.x and NuChess aren't around even >in their grotesque Fortran. Where did Slate and Atkin disappear to anyway? > >Stuart
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.