Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hash hit != Hash full

Author: Sune Fischer

Date: 11:48:25 06/19/04

Go up one level in this thread


On June 19, 2004 at 14:26:35, Gian-Carlo Pascutto wrote:

>On June 19, 2004 at 14:07:31, Sune Fischer wrote:
>
>>On June 19, 2004 at 13:45:59, Gian-Carlo Pascutto wrote:
>>
>>>On June 19, 2004 at 13:39:47, Sune Fischer wrote:
>>>
>>>>When having to choose though, I'd have to say that I just don't pay very much
>>>>attention to the different counters during a game, except perhaps the nps.
>>>
>>>IMHO one has to make this number useful by reporting a useful value
>>>to the user :)
>>>
>>>Wouldn't you agree that when it's pegged at 99.9% constantly it would
>>>be a good idea to give more hash, or conversely that when it's constantly
>>>at 1% that maybe more memory could be used for tablebase caching?
>>>
>>>There is a chance to give some good information here.
>>
>>Absolutely, as usual it's a few oddball situations that's bothering me.
>>
>>E.g. suppose I analyze a position to fill 99%, then go one move forward and let
>>it fill to 30%, and then go back again.
>>
>>Now the max fill will be 30%, because the remaining 70% has the same flag so the
>>counter will not increment when hitting those.
>>
>>Perhaps a flaw I can learn to live with, I guess there is no perfect method.
>
>Deep Sjeng doesn't have any problems with this. When you go back and forwards
>again, the counter won't consider the positions the same. However, each time
>and old position gets a hashit (which will happen a lot!), it will get promoted
>to the current search.
>
>So you'd effectively have the same efficiency and hash statistics still
>work correctly.

Right okay, using an age not based upon the ply in the game would solve it.

I'll consider using it :)

-S.
>--
>GCP



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.