Author: Eugene Nalimov
Date: 12:39:17 05/05/99
Go up one level in this thread
On May 04, 1999 at 12:54:33, Peter Fendrich wrote: >On May 04, 1999 at 12:05:35, Steven J. Edwards wrote: > >>On May 03, 1999 at 21:21:55, Christophe Theron wrote: >> >>>On May 03, 1999 at 18:27:44, Steven J. Edwards wrote: >> >>>>The freely distributed tablebase generator source is available at no cost >>>>via e-mail to any interested parties. It is sent as a uuencoded gzip >>>>comrpessed tar archive, so you will need the uudecode, gunzip, and tar >>>>utilities (or their functional equivalants) to get to the source. The >>>>file is just under 50 Kbyte long. The program is written entirely in >>>>ANSI C and had run under Mac, Linuc, and Windows. >>>> >>>>Send e-mail if you are interested. >> >>>Is it designed exclusively for Crafty? >> >>No. The tablebase generator is a separate program. The files it produces >>are simple binary data files that may be used by any compatible program >>on any platform. >> >>>Is the source code for accessing the tables avalaible? >> >>The Crafty source distribution contains working sample code for accessing the >>tablebase files. >> >>>Does it work under DOS too? >> >>Yes. It is a simple "console" style application. >> >>-- Steven > >How are the generated files compared to the E. Nalimov type? >I suppose bigger - any ratio? >Faster? >//Peter I used SJE's TBs as reference point, and their existance greatly helped me during debugging. Nevertheless, my code is 100% original, except a several indexing functions that are used only in 'SJE compatibility mode'. Sum of all 4-man TBs: SJE's TBs are greater than mine (uncompressed) by ~80%. Sum of all 3+2-man TBs: SJE's TBs are greater than mine (uncompressed) by ~120%. There is the article by Ernst A. Heinz in the last issue of JICCA about different indexing schemas with brief reference to my work, and we are planning joint article in the "Advances in the Computer Chess 9". Compressed TBs are ~3.8 times smaller than uncompressed, so when compressed, mine TBs are ~8 times smaller then SJE's. Based on the experiments with Crafty, my probing code is significally faster due to TB caching. Sorry, do not remember exact details. You can download source of generator, probing code, and instructions how to integrate probing code into your chess program, from Crafty web site ftp://ftp.cis.uab.edu/pub/hyatt/TB/tbgen.zip (source), ftp://ftp.cis.uab.edu/pub/hyatt/TB/tbexe.zip (win32 executables). Generator and probing code was succesfully compiled and run at x86/Alpha NT (my develpoment environment), x86 Linux, Alpha Digital Unix, and Sun Solaris. I hope that the last version can run at Mac, too, but cannot be sure. In any case, I can help with Mac port. Generator and probing code are 'Crafty-neutral' as well as 'endiannes neutral'. As far as I know, probing code was incorporated into several amateur programs and into at least 4 commercial programs. First of them - CSTal2 - was already released (unfortunately, without support of compressed TBs; planned in CSTal3). I estimate that it takes 1-3 hours to incorporate probing code into the chess program. Eugene
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.