Author: Alessandro Damiani
Date: 09:52:54 11/22/02
Go up one level in this thread
On November 22, 2002 at 12:19:42, Sune Fischer wrote: >On November 22, 2002 at 11:58:34, Alessandro Damiani wrote: > >>On November 22, 2002 at 11:37:56, Sune Fischer wrote: >> >>>On November 22, 2002 at 07:42:51, Gerd Isenberg wrote: >>> >>>>For move generation, or if i really need disjoint attack sets, i currently favor >>>>attack bitboards for a single sliding piece as well for a knight. That is what >>>>my program does before with rotated bitboards and now with dumbfill and >>>>Kogge-Stone mmx-routines. >>> >>>There is another thing which bothers we with disjoint attack boards, namely that >>>they are _disjoint_. >>> >>>Some parallelism seem to be lost here, like how do you quickly detect a knight >>>attacking a certain square if your knight attack boards are split into 8? >>> >> >>Those 8 knight attack bitboards are only used for move generation. Each time a >>move has been searched the to-square is removed from such a bitboard. >> >>To do what you say the flood-fill attack detection functions are used. > >You still need that NxQ pattern and all the others in your move ordering. > >I suppose one could just OR them together. > We can generate NxQ captures one at a time. I didn't think further to see if SEE can be done more efficient with Steffan's method. Steffan? You wrote someting you name SME? Does it include a SEE? Alessandro
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.