Author: Guido
Date: 10:12:43 11/07/01
Go up one level in this thread
On November 06, 2001 at 08:19:04, Simon Finn wrote: >On November 05, 2001 at 21:08:43, Slater Wold wrote: > >>Geez. That is incredible. The decompressed 5's are probably 20GB. (Compressed >>they are 7GB.) >> >>All that computing power, to create the krrrrk in 4 months. How exciting. :) > >Where do you get the figure of 4 months for krrrrk from? > >Due to the internal 24-way symmetry (4! = 24), krrrrk should be >slightly smaller than krpkr. The longest mate is much shorter >so generating krrrrk should require fewer iterations than >generating krpkr; this counter-balances the increased branching >factor. > >Generating krpkr on an 1333 Athlon took roughly 2 days; >I would expect krrrrk also to be do-able in a couple of days >on a decent machine. > >Or does Nalimov's code process each krrrrk position 24 times >on each iteration? > >Simon Solving krrrrk and kqqqqk (more exciting with four queens :-)) with my program takes a little more than 4 and 2 hours respectively by a 600 Mhz PC. But I solve this ending using only two bits for position (win, loss, draw and illegal position), so the need of memory is only 128Mb, instead of 512 Mb for generating the tables with one byte for position (result + number of moves). Nevertheless I think that for_these_endings the cpu time shouldn't be very different between 2 bits or one byte per position cases as the number of cycles are probably the same, while for other endings cpu time could be quite different. Ciao Guido
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.