Author: David Rasmussen
Date: 02:19:19 01/27/03
Go up one level in this thread
On January 26, 2003 at 17:34:03, Walter Faxon wrote: > >Hi, All. > >In Crafty at least, there are sometimes reasons for scanning from a particular >end. For example, finding the most advanced passed pawn for black requires a >search from the opposite end from finding the most advanced passed pawn for >white. As of version 18.10, all such side-dependent code is isolated in >MOVGEN.C. (And of course, ranks must have the minor order in the bitboard for >this to work: e.g., a1-h1 are contiguous bits.) > >So: for Crafty on any particular architecture, one would want three routines: >FirstOne(), LastOne(), and say, NextOne(), using the faster/smaller/"better" of >the two when it doesn't matter which way you scan (like when adding up >feature-weights). > That was what I meant. I know that it is sometimes important to scan in a specific order, but if we can have a speedup when it isn't, we should. I don't know if a faster one can be made with this looser criterium. Also, often we clear a bit after finding it, so if it can be done faster in one operation (finding (first,last or some) and clearing), then that could help too. /David
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.