Author: Robert Hyatt
Date: 07:31:41 07/07/03
Go up one level in this thread
On July 06, 2003 at 19:41:56, Federico Corigliano wrote: >In my chess program I'm using a single transposition hash table. When I will >store info it, and already are info on it for the position, I check if the info >in the table has a smaller depth. If this is correct, I replace it with the new >info. >I saw that other people use two transposition hash tables: one "always replace" >and other with a "special replace scheme". >What's the difference between the "special replace scheme" and the system that I >use? >How much in the improvement using two hash tables, one with "always replace"? > >Thanks If you use one table, and when you do a "store" you only check one position, you can lose important information. With two tables, you have two choices on what to replace, which is better. Of course, with one table you could do the same thing (as I do in Crafty now), and with one table you _could_ look at as many positions you want to try to minimize your loss when you overwrite something. The better your replacement strategy, the smaller your overall search tree. However, if you get too clever, the cost of your hash probes will exceed the savings in tree size and the net change is a loss.
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.