Author: Anthony Cozzie
Date: 12:33:36 08/31/03
Go up one level in this thread
On August 31, 2003 at 14:55:58, Vincent Diepeveen wrote: >On August 30, 2003 at 18:05:01, Anthony Cozzie wrote: > >>On August 30, 2003 at 17:05:45, Vincent Diepeveen wrote: >> >>>In REBEL as we can read from its homepage it is also 1 instruction. >>> >>>Because Ed doesn't know the difference between C and BASIC, >>>i'll write down the code for REBEL here in C (of course rebel-engine is 100% >>>assembly so 'decompiling' is not happening here) >>> >>> attackers = MyAtt[sq] & 7; >>> >>>He that looks very similar to what i do in DIEP! >>> >>>Now fritz: >>> >>> attackers = MyAtt[sq] & 7; >>> >>>He HOW SIMILAR that looks to rebel? >>> >>>Must i go on? >> >>how is this incompatible with bitboards? I don't think there is some law saying > >It is incompatible because generating them with bitboards is going to be 6 times >slower than without and that isn't a joke. > >Of course you can use a crippled bitboard version, so converting when having a >bitboard with moves, the moves to a square and then one by one nail them in into >the attacktable. > >Then you can limit the speed loss to a factor 3 perhaps. > >>you can't have bitboards *and* attack tables. I am giving very serious thought >>to including Ed-style tables in Zappa. Crafty's king-safety code is (imo) >>inadequate. With pieces like bishops, it trivially doesn't matter if the bishop >>is at B1 or D3, it still is attacking the king's field. Maybe at some point >>Zappa's bitboards will simply get phased out of existence, but I doubt it. > >I feel the crucial point is what type of code you use to generate moves, this >will determine whether your program is a bitboarder or not. Because if you do >that with bitboards, then your scanning code will be all bitboards too most >likely. Even more likely you'll hardly have such scanning code (which the >commercial engines have bigtime). > >>They >>are simply too convenient for some things. > >There is many things which are said to be quick in bitboards which i already do >for 8 years now in a different way just as quick, if not quicker. > >But by far the real reason why bitboards seem quick to some, is simply because >they do not have all that fancy stuff the commercial guys have. > >If you don't have something you don't waste system time at it. with 99% sureness >you can say then also that they won't show up at a world championship either. > >Last bitboarder i remember joining a world championship in person has been >somewhere London2000 - James Robertson. > >Note that if i remember well he had written large parts in assembly. > >What a waste of time... > >>The rotated bitboard masks are L2 yes, but the others (like the k-p square) are >>probably L1. >> >>anthony I don't think using bitboards and having a big eval are mutually exclusive. Most amatuers just aren't interested in eval, because its a lot of work, and it doesn't show up on test suites. Also, I haven't found a use for attack tables outside of kingsafety and SEE, and I have quick ways to do both of those with bitboards. anthony
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.