Author: Christophe Theron
Date: 22:49:47 06/16/03
Go up one level in this thread
On June 16, 2003 at 23:00:57, Bob Durrett wrote:
>On June 16, 2003 at 21:27:07, Ricardo Gibert wrote:
>
>>On June 16, 2003 at 19:48:58, Bob Durrett wrote:
>>
>>>
>>>I do not know whether or not this is a new idea. Whether it is or not, does it
>>>have any merit?
>>>
>>>Idea: Variable Hash Table size
>>>
>>>Someone said it is possible to have a hash table size too large, and sometimes
>>>too small.
>>
>>
>>But I don't think they were implying this was a problem *during* a game, but
>>rather *between* games such as for example a 1-minute bullet game using a
>>reasonably sized table followed by a 40/120 game that could make good use of a
>>larger table.
>>
>>A variable size hash table is not for free. You retain the O(1) performance
>>metrics, but it will still cost you some speed. It also will make the program
>>more complex and who needs that when there is really no payoff?
>
>Thanks for your comments.
>
>My interest in a chess engine is solely for using it as a tool to help with
>post-mortem analyses of games. If there is no payoff in that situation, I am
>not interested. Currently, I set my hash table size to the largest allowable
>integer power of two. That seems right?
>
>Bob D.
If you use Chess Tiger that's perfectly fine.
Actually even if you were playing bullet games, this setting would still be
optimal for Chess Tiger.
That's why my program does not need the idea you suggested: for Chess Tiger, the
bigger is always the better (unless you start to reach a size that would force
the OS to swap some parts of the HT to disk, but I guess it was not your point).
Christophe
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.