Author: Vincent Diepeveen
Date: 15:14:46 11/19/02
Go up one level in this thread
On November 19, 2002 at 15:08:13, Daniel Clausen wrote: please mention me 1 bitboard program with a big eval. *NONE*. To me bitboards seems something for people who are no good programmers, because they can cut'n paste from crafty and go further with that. Optimizing gnuchess or gerbil or whatever to something real fast for your needs is way more difficult of course than starting with something that's working and written out in detail. Usually people also cut'n paste the SEE and qsearch from crafty then and they have something much better than they can produce in a lifetime most likely. That's the only attractive things from bitboards IMHO for several authors. And as long as they don't improve the evaluation a lot it remains like that. If on the other hand you look to what representation the good programmers go for, the picture is real clear. this has nothing to do with religion but with objective speed differences. My move generator without inline assembly and with general code for both sides, it is 2 times faster than crafty at any x86 processor. That's *objective* measurements. My SEE is better than the one from crafty, picking up more than Crafty does in the SEE. Very objectively provable. The list goes on and on. Most important thing however IMHO is that the source from crafty is free. If mine was free, everyone would start with DIEP and go further from there. I'm 100% sure of it. We saw this before. When GNUchess was the strongest freely available source code, people started with that crap. I wrote nearly every byte of my move generator. *every* byte. It took me years to make a fast generator. Not everyone is that great. >On November 19, 2002 at 14:04:42, Vincent Diepeveen wrote: > >[snip] > >>i have posted some months ago and another few months before that loads >>of examples with regard to evaluation. >> >>If you browse some in the search you will find it. > >I'm aware of that. But I can't remember that you scientifically proved that >"bitboards are worse to implement a good eval than 0x88". (or any other board >representation) > >Actually it would be a rather stupid claim to make because there's really no way >you could prove that. (on the other hand, religions make use of the fact that >their claims are not provable/disprovable ;) > >Just posting some examples where 0x88 is better than <another board >representation> is not a proof. In fact I'd be surprised if _your_ evaluation >would be easier/faster to implement with bitboards than with 0x88, as it would >mainly show that you didn't make use of the advantages of your chosen board >representation. > >While there are clearly inferior board-representations (like storing the board >internally as a BMP-file ;), generally the art is to find the advantages of the >chosen representation and make use of them. (that's not only true for chessboard >representation but for many other things) > >Sargon
This page took 0.01 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.