Author: Andrew Dados
Date: 10:35:43 05/15/99
Go up one level in this thread
On May 15, 1999 at 13:23:33, KarinsDad wrote: >On May 15, 1999 at 11:36:18, Andrew Dados wrote: > > >> >> Ok.. let's summarize.. Your format ends up with 22 bytes and does not >>support all legal positions.. Now let's check way easier schema: >>64 bits locations = 8 bytes >>32x4 bits piece+color flag =16 bytes >>50 move + ep+ castling rights+ side to move = 2 bytes >>total gets to 26 but *any* 32 piece position is supported and way easier/faster >> to code.. >>As usual - trade off.... >> >> -regards - >> Andrew > >You forgot move by repetition, 2 more bits. So, a total of 27 bytes for your >schema. > >Our schema is harder/slower to code. You are absolutely correct. However, the >decompression will not be THAT much slower than the schema you mentioned, so >once it is coded and debugged, the main difference is the smaller size. > >The original post asked for the smallest representation since for the reasons in >which you would wish to use this (such as an opening book), you want tight >compression, not ease of accessing. The hard disk access is WAY slower than the >decompression in memory if you are accessing a lot of hard disk simultaneously. > >Our schema can support all legal positions in 25 bytes. > >Our schema can support all legal positions that would every show up in a real >game in 22 bytes (and if you were concerned, you could increase it to 23 bytes >and never have to worry about it). > >Our schema can support an opening book where you do not need the 50 move rule in >21 bytes. > >I doubt our schema is the best, but it comes closer to the original post of the >tightest compression possible than the simple schema you mentioned by several >bytes. Also, from what I can tell, this problem has been solved before by others >to the level at which we did it (i.e. we just reinvented the wheel the exact >same way, so we cannot take credit for it). > >KarinsDad :) ok.. so 25 bytes... If I now define 2 unused pieces (coded in 3 bits) as 'eppawn' and 'castle rook', and also use 'eppawn' somewhere for side to move (impossible) to flag repetition I also end up in 25 bytes fixed size.. means in all hash applications, where all it counts is maximum size (cause of fixed record size) we end up even :). While coding such a schema versus yours is *way* less pain... always lazy I was.... -Andrew-
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.