Author: Jason Williamson
Date: 16:23:50 12/01/01
Go up one level in this thread
On November 30, 2001 at 21:18:59, John Merlino wrote: >On November 30, 2001 at 18:36:58, Jason Williamson wrote: > >>Surely that time could be improved? Or is it still memory issues as the major >>problem? > >Definitely memory issues. The import that Chessmaster uses (which, admittedly is >quite ancient and nobody really wants to touch it) is rather slow. But the >problem is the memory requirements. A lot of time could be spent on optimization >and so forth, but it is hard to justify it with regards to the schedule, >considering that the feature is probably used by only 0.01% of our users (if >that!). > >During import, each node requires 43 bytes of data. This adds up very quickly >when you're talking about 40 moves each of 140,000 games. Worst case scenario >(which isn't actually possible, as there have to be SOME repeated moves) would >be over 240 million bytes of RAM required. > >I did spend a fair amount of time (i.e. a couple of days) trying to improve >things, and this was the result. The original code (written long before I got on >the scene) required 92 bytes of data per node. But it wasn't a problem because >the format limited a book to a maximum of 64K moves. So the maximum memory usage >for the old format was only about 6MB -- not a big deal. > >If I felt like spending a couple of weeks, things might be improved noticeably. >Unfortunately, I don't feel like it! (and don't have the time even if I did feel >like it). > >Hopefully, the quality of the new book will make it so the opening book editor >is used even less! > >jm Yeah, if the new book is good, there won't really be much need to wreck it with random PGN games. :D For most engines I use the book it comes with, ie, Fritz gets its Fritz book and so on.
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.