Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Let's analyze move 36

Author: Robert Hyatt

Date: 09:51:44 10/08/97

Go up one level in this thread


On October 08, 1997 at 07:08:01, Amir Ban wrote:



>Don't believe it. The fact that hash collisions occur does not mean that
>it affects the PV, and if it does, it is a really freak accident. I do
>48-bit hashing with almost no validation. If you wait for my program to
>fail because of that you will get old in waiting.
>

this is not necessarily true.  Several of us, in a long thread in
r.g.c.c a couple of years ago, very carefully measured the number of
hash collisions produced using a 32 bit, 48 bit, and 64 bit hash key.
32 bits is totally hopeless.  48 bits was better, but still produced a
large number of collisions at high node rates.  64 bits produced a
*significant* number of hash collisions as well.  These were all run on
machines that were then searching 20-30K nodes per second, except for me
(and the 64 bit numbers) where I ran the test on a C90 at 500K nodes per
second or so.)

We are getting far more collisions than you imagine I suspect, based on
the numbers from Crafty, ZarkovX, I believe Ed contributed some results,
and I don't know who else was involved.  To think that multiplying by
2000 is really like removing 11 bits from the hash signature is a
sobering thought.  It is likely that they are on the fringe of seeing
bad things happen, particularly when they search for 20 minutes at a
pop.




This page took 0.02 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.