Author: Uri Blass
Date: 03:46:57 01/09/03
Go up one level in this thread
On January 09, 2003 at 06:36:22, Uri Blass wrote: >On January 09, 2003 at 06:04:27, Tony Werten wrote: > >>On January 08, 2003 at 18:19:39, Uri Blass wrote: >> >>>On January 08, 2003 at 17:22:40, Tony Werten wrote: >>> >>>>On January 08, 2003 at 16:25:44, Russell Reagan wrote: >>>> >>>>>On January 08, 2003 at 14:31:10, Ed Schröder wrote: >>>>> >>>>>>Yes Gerd, this is probably most true. I never had the time to put energy in the >>>>>>bit-boards technique. Do you have any idea about the number of clocks needed to >>>>>>extract the hanging pieces from a position using bitboards? >>>>> >>>>>I wrote a program using the same algorithms that Gerd uses for generating >>>>>attacks, but mine are written in C instead of using MMX like Gerd, so mine are >>>>>going to be slower than his. In any case, the engine is somewhat slow. It gets >>>>>around 70knps or maybe a little more on my PIII 733 MHz machine (compare with >>>> >>>>I don't want to bring you down but that is kind of slow. >>>> >>>>XiniX, while generating Ed's attackboards and doing full evaluation and >>>>everything else still gets 400Kn/s in the middlegame on a amd2000 >>>> >>>>And that's only with generating the attackboards, not actually using them to >>>>speed up stuff. >>>> >>>>Tony >>> >>>Movei probably makees 300-350 Kn/s but based on your definition >>>of nodes it is probably more nodes. >>> >>>Movei generates millions of legal moves and I know from previous discussion that >>>you consider part of the moves that it generates as nodes when I do not count >>>them as nodes. >> >>Actually, I don't either. The discussion was about futility pruning. I have the >>opinion that if you do all the work for a move and then prune it, you should >>count it as a node. >> >>I don't prune in XiniX other then nullmove. So my 400 should be 400 to you too. >> >>Tony > >I do not know what is all the work because for me making the move is also some >work because I update my attack tables and later I need to generate the legal >moves to calculate exact evaluation. > >Uri To be more correct today I need only to know the number of the moves for exact evaluation but I may want to extend based on the result of exact evaluation so if I do a special function only to count the number of legal moves that it may be a waste of time in that case and in any case I believe that there are better ways to evaluate mobility than counting the number of legal moves(like Ed I have a table and the evaluation is not a linear function of the number of legal moves and I also use the number of legal moves of the opponent based on previous move) Uri
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.