Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fitz5 and hash for once

Author: Georg Langrath

Date: 12:44:16 05/31/98

Go up one level in this thread


On May 31, 1998 at 12:39:59, Robert Hyatt wrote:

>On May 31, 1998 at 11:45:02, Georg Langrath wrote:
>
>>q5k1/4Q3/5r2/5P1B/6PK/8/8/8 w
>>
>>This earlier showed "draw-position " should be suitable to proof
>>hashtables in Fritz5, I thought. The resullt was  illogical and
>>difficult to understand for me. I have Pentium 133 and 32 MB memory . I
>>noticed the time when the computer understood that the position is draw.
>>When I repeat it, I got the same result. Has my Fritz tasted alcohol?
>>
>>Hashmemory	Time
>>
>>16448		46s
>>14400		11s (!)
>>13376		1m 58s
>>12352		49s
>>11328		50s  Here Hashmemory is "Full" first time.
>>10304		14s (!)
>>9280		1 m 55s
>>8250		1m 7s
>>7232		26s (!)
>>6208		14s (!)
>>5184		45s
>>4160		16s (!)
>>3136		11s (!)
>>2112		1m 11s
>>1088		5m 51s
>>64		>10 minutes
>>

Georg
>
>
>this has been explained many times, and is based on random chance.
>Positions get overwritten or not overwritten based on the random values
>used to produce the hash signature, and the size of the hash table.
>Often,
>overwriting something will speed the search up, but equally as often,
>overwriting something will slow it down.  It is not a phenomenon that
>lends itself to precise measurement, because of the randomness of the
>hash signature.  But it is simply a fact of life in computer chess or
>any other place where randomness is used...
>
Thanks. Sorry that it has been explained several times, but as it comes
new members there is always a risk for repeats.

Georg
>
>




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.