Author: Vincent Diepeveen
Date: 05:19:31 10/01/02
Go up one level in this thread
On September 30, 2002 at 10:41:30, Tom Likens wrote: >On September 30, 2002 at 04:08:02, Vincent Diepeveen wrote: > >[snip!] > >>I'm no exception to that. > >I take the 5th here :) > >>But with regard to speed: in assembly the 0x88 is very fast. Otherwise >>much easier to make than all this is using the gnuchess 4.0 datastructure >>(don't get the raped 5.0 versions which are bitboards, but the >>'int board[64]' stuff from before that). It's very easy and can be speeded >>up by good programmers a lot. >> >>I really would go for that gnuchess 4.0 stuff, simply because your thing >>gets a lot easier to write in other parts of the story. > >It's interesting but I still use gnuchess 4.0pl80 to test changes to my program. >Especially, now that 5.x has branched off, using the 4.0 series gives me a >stable benchmark to compare against. Note, many of the commercial >programs are *not* bitboard driven, so they are no magic panacea. The one >advantage you will have using bitboards is that you have access to the source >code of a strong program that uses them (Crafty). > >Just some general advice, no matter what data structures you end up using >you should build in a number of ways to test that your engine is solid. If you >decide to go with bitboards I would get the non-rotated version going first. >Bitboards are hard enough initially, without rotating everying 45 or 90 degrees. add to that, that you need to write down for every piece everything in bitboards. you need loads of inline assembly for nearly any architecture to keep going and you need a 64 bits processor to not be 2 times slower than a non-bitboarder. also crafty is perhaps a great program to learn from for beginners, but for advanced computerchess it's obviously outdated. therefore i'm not using bitboards. >regards, >--tom
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.