Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: hash collisions

Author: Carmelo Calzerano

Date: 05:53:41 02/02/01

Go up one level in this thread


On February 01, 2001 at 20:34:32, Vincent Diepeveen wrote:

>On February 01, 2001 at 12:54:53, José Carlos wrote:
>
[snip]

>>  I'm sorry to get into your argument, but I think you're missing that,
>>what he means is "what to use as an index for the table where you store
>>the positions.
>>If I'm not wrong, he wants to _store_ the whole position but use zobrist to
>>calculate _indexes_ to those stored positions.

That's exactly what I meant: use zobrist (or something similar) _just_ for
indexing, not as hash signature

>>And that makes perfect sense to me, because if you use the n-less
>>significant bits of the signature (as we do in chess) you're gonna get
>>the same index for a buch of very similar positions.

Exactly

>Yes a very quick way to store is 32 bits zobrist hashing and
>a true store of the position!

Looks like finally we agree :-)

>Another and quicker way is to XOR the 63 bits of the true storage
>internal with each other, though i'm sure that zobrist gives
>a better division over the hashtable as that...

I agree; this way index distribution would be far worse than
using zobrist (although still much better than simply masking bits)

>But one can invent another zillion ways of indexing them :)

Of course. But if you find any better way please let me know...
:-)

Bye,
Carmelo



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.