Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hash codes (test results)

Author: Gian-Carlo Pascutto

Date: 02:08:20 11/16/01

Go up one level in this thread


On November 15, 2001 at 22:50:34, Ricardo Gibert wrote:

>On November 15, 2001 at 07:24:34, Gian-Carlo Pascutto wrote:
>
>>On November 14, 2001 at 12:33:09, Ricardo Gibert wrote:
>>
>>>How many of the test positions were endings?
>>
>>Should be about 20-30%
>>
>>>What were the sizes of the 32-bit key hash tables and 64-bit key hash tables?
>>
>>Both three seperate tables of 1 000 000 entries each.
>>
>>>How large are your hash entries?
>>
>>16 or 20 bytes
>
>If understand you correctly, you used the same number of entries for both, but
>this discounts one of the advantages of using 32 bits. To make the test fair,
>you should have them use the same amount of memory. Something like 1,000,000
>entries for 32 bit and 800,000 entries for 64 bit. In that case, the higher
>probability of a collision for the 32 bit table will be ameliorated somewhat by
>being to fit more entries in the same amount of space.

True, but my goal was not to demonstrate the superiority of 32-bit hashes,
but to provide counterevidence to the myth that 32-bit hashes are 'not enough'.

--
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.