Author: Robert Hyatt
Date: 07:16:44 12/04/99
Go up one level in this thread
On December 04, 1999 at 07:55:40, Alessandro Damiani wrote: >On December 03, 1999 at 19:27:50, Nicolas Carrasco wrote: > >>What bitboards method is faster : Chess 4.5 "old" or rotated bitboards? >>I know that it depends how you use it, but I would like to hear an answer. >> >>Thanks >> >>Nicolas >>http://www.puntadeleste.to/uruchess/ > >In Fortress I don't use "rotated bitboards". Of course, every method is a >rotation, but the one I use works differently. It is extremely simple and does >not need to scan the board. But I don't know if, for a 64-bit machine, it is >faster than "rotated bitboards". I don't know if Bob has made some tests. > >I sent the source code to Bob, because I decided to make it Freeware. Bob asked >me to publish it in the ICCA journal, but I had to stop writing. It takes me too >much time. > >If there is enough interest here I will post it. Remember it is Freeware. > >Alessandro I think your approach is faster on 32 bit machines, because you don't keep the 'rotated' bitmaps in a 64 bit value, you keep each rank/file/diagonal in a separate 'char' array element, which means you avoid the shifting/ANDing that I have to do to extract this. However, on 64 bit architectures, the traditional approach seems better as shift/AND are 1-cycle operations, and the entire bitmap fits into one register. shift/AND is much faster than computing an array index and then poking into the 16-byte array... But for 32 bit (and less) machines, avoiding some of the 'work' is a good idea..
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.