Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A question about pawn hash tables

Author: José de Jesús García Ruvalcaba

Date: 10:40:39 09/14/00

Go up one level in this thread


On September 14, 2000 at 12:31:07, Carlos del Cacho wrote:

>On September 14, 2000 at 09:58:35, Andreas Herrmann wrote:
>
>>On September 14, 2000 at 08:38:31, Andrew Williams wrote:
>>
>>>On September 14, 2000 at 05:25:32, Andreas Herrmann wrote:
>>>
>>>>On September 13, 2000 at 17:12:07, Andreas Herrmann wrote:
>>>>
>>>>>Hello,
>>>>>
>>>>>i want to include pawn hash tables in my chess program holmes, but i don't know
>>>>>how i have to calculate the hash index and the hash key.
>>>>>I think the hash index of the position will only calculated with the pawns on
>>>>>the board like the following,
>>>>>
>>>>>for (pos=A1;pos<=h8;pos=pos+1) {
>>>>>  if (figure[pos]=pawn) {
>>>>>    PawnHashIndex = PawnHashIndex ^ RandomIndex[pos].index;
>>>>>  }
>>>>>}
>>>>>
>>>>>is this right??
>>>>>
>>>>>How can i get a good hash key? What have i calculate with xor?
>>>>>
>>>>>Thanks for each information.
>>>>>
>>>>>Andreas
>>>>>
>>>>>
>>>>>www.wbholmes.de
>>>>
>>>>Hello again,
>>>>
>>>>the links from Ricardo are a good information source, but i found nothing about
>>>>pawn hash tables, only normal hash table and that's not new for me.
>>>>
>>>>My question is how can i get a good hash key for collision check in the pawn
>>>>hash table.
>>>>Is a XOR combination with a random figure enough? Or is it not so difficult than
>>>>i am thinking?
>>>>
>>>>Andreas
>>>
>>>Yes. I've always used a 64-bit integer derived in just the way you describe.
>>>I don't know if 32-bits would be sufficient, as I've never tried it.
>>>
>>>Andrew
>>
>>Ok a 64-bit integer is of course much better against hash collision than a
>>32-bit. I will try to implement it with a 64-bit interger.
>>
>>thanks
>>Andreas
>
>I've been using 32 bit nums since I first implemented it in pepito a few months
>ago, and I haven't had any problem with collisions yet. I guess the pawn
>structure changes so slowly in search that the chances of getting a collision
>are really small unless you get to incredibly deep lines (with lots of pawn
>moves...  I don't think that's possible).
>
>Carlos

	Can you tell us more about Pepito? I assume it is your chess engine. Any chance
of seeing it in tournament action soon?
José.



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.