Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: HashTable upper_bound worthless ?

Author: Larry Griffiths

Date: 19:09:24 02/14/01

Go up one level in this thread


On February 14, 2001 at 14:06:03, José Carlos wrote:

>On February 14, 2001 at 12:37:52, Larry Griffiths wrote:
>
>>On February 14, 2001 at 10:11:53, Carmelo Calzerano wrote:
>>
>>>On February 14, 2001 at 09:35:58, Larry Griffiths wrote:
>>>
>>>>Carmelo,
>>>>
>>>>I will dink around with the hash table some more and see if I can find a bug
>>>>somewhere.
>>>>
>>>>Is this Fine 70?
>>>>
>>>>[D]8/k7/3p4/p2P1p2/P2P1P2/8/8/K7 w - -
>>>
>>>Exactly ;-)
>>>
>>>Bye,
>>>Carmelo
>>
>>Ok Carmelo,
>>
>>I did find a bug in my hash code and am getting upper_bound cutoffs now.
>>I owe you one since I was about to give up on the hash table for a while.
>>I ran the Fine 70 position using a 16meg hash table.  My hash tables are
>>28 bytes in size.
>>
>>My program goes 17 plys in 18 seconds without the hash table,
>>and 19 plys in 6 seconds with the hash table.
>>
>>I noticed that Hardly any of the 16meg hash table was used.  I am only using 1
>>slot per hash key, and maybe if I add/look at the next 8 slots I can get some
>>more entries in the hash table and go deeper on plys.
>
>  I don't know what you mean with the slots thing (though I figure), but I think
>you may have a problem with the hash codes, because you're doing too slow in
>this problem, and don't even fill 1 Mb of the hash...
>
>  I may be wrong, but I think your hash codes are not different enough for
>similar positions.
>
>  José C.
>

Thanks for the information Jose.  I generated my piece-square tables using
random numbers and the hash-keys in the tables might not be different enough.
I am using 64-bit hash-keys as 32-bit hash keys were producing lots of instances
where different positions equated to the same hash-key.  I think this caused my
program to make some terrible moves so that is why I went to 64-bit hash-keys.
I will try using some different hash-key generations schemes and then see what
percentage of the hash-tables is used.

Larry.



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.