Author: martin fierz
Date: 15:48:20 05/10/02
Go up one level in this thread
On May 10, 2002 at 17:25:49, Alvaro Jose Povoa Cardoso wrote: >Hi Martin, >I would like to know your opinion on LZSS for EGTB. >LZSS is extremely fast on decompression. >I tested a 104Mb pgn file with LZSS and it compressed to 35Mb. >I know pgn files is not the intended purpose but do you think it has good >aplicability for EGTB? > >Best regards, >Alvaro Cardoso hi alvaro, i don't even know what LZSS is :-) i'm currently very busy working on my program for the las vegas tournament, so i'm not trying for better database compression at the moment. i have compressed my database from about 37GB to 4.3GB; but it is a lossy compression scheme, removing all positions where a capture can occurr from the database. this makes a lot of sense because these are typically positions with "random" values. the chinook db is compressed to 5.6GB, both mine and the chinook db use run-length encoding. at the time chinook was programmed (about 1992-1996), schaeffer wrote that RLE was the only viable method for db compression - as it was fast in decompression. since then, processor speed has increased much more than harddisk speed, so maybe today a scheme with longer decompression time but better compression would do much better. once i need disk IO in my program, it just dies - it typically does 1000kN/s in an opening position, and that goes down to 500kN/s in an endgame position, where it has all the db blocks in memory. if it has to read from disk, it can get as slow as 5kN/s! winzip compresses my compressed database down to about 2.5GB. so that's about as good as it can be. i definitely plan to look into other compression methods after the vegas tournament is through. for example, the compression for chess EGTB with datacomp.exe is about 30-40% better than mine - already quite close to the winzip values. i don't know what it uses aloha martin
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.