Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Bitboard vs Offset again ! now after Rybka

Author: Gerd Isenberg

Date: 13:36:42 01/20/06

Go up one level in this thread



>I have not read any formal BB technique, too tough. So I just go ahead armed
>with x & -x, x & x-1, x - y > 0 to get intervening bits, bits_E/LE/G/GE etc..
>
>I now cannot see how any BB engine can be efficient w/o incremental updates of
>all (sliding) attack maps of the b/r/q. The complete attack info is an
>incidental good bonus. I can only see that to extract Q-moves, we need the
>complete q-attack map. If we generate the Q-attack map anew at every node, we
>need to manipulate all 8-rays. With incremental, at most 2-rays or none, need
>updates. My 1-step-ahead IQ asks _how-to-be efficient?


With rotated, four lookups for queenattacks with shifted or already raywise
updated occupied states and square. Not that expensive.

It might be very implementation depending - the number of nodes you need queen
attacks (for both sides) and the number of nodes you don't need them or only for
the side to move.

>
>May be the reason why Rybka quickly exchanges it's queen is not to deal with
>8-rays. Vasik would likely go to trade all R-R, B-B, so nothing sliding to
>worry. Then his Knight = 330, Bishop = 310, etc...

Hehe - i don't think so.

Regards,
Gerd

>
>Best Regards
>Rasjid



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.