Author: Johan Hutting
Date: 09:42:23 09/06/03
Go up one level in this thread
On September 06, 2003 at 07:44:48, Uri Blass wrote: >Is there an important reason to use more than small number of entries(say 2^14 >entries for pawn hash tables). More stores, more hits. However, the benefit may drop rapidly. This is something you can test yourself. > >How much speed improvement do you get when you compare 2^14 entries for pawn >hash and 2^17 entries or more entries? Again, this is something you can test yourself. You want to know how much it improves Movei, not engine XYZ. > >Is it something constant or something that is dependent on the main hash tables? I'm using a constant value. Unlike eval and transposition hash there are not that many possibilities, so... :) YMMV, test test test Uri. > >What is the percentage of hash hits for pawns as a function of the number of >entries? Test :) Or alter crafty or any other open-sourced program with pawnTT. I'm not interested in getting the perfect pawnhashsize, it's too dependant on many factors and my current number works fine (crap for start position, 94+% in most others) so far. Regards, Johan
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.