Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hash Table Collisions

Author: KarinsDad

Date: 08:11:24 04/13/00

Go up one level in this thread


On April 12, 2000 at 22:59:22, Tom Kerrigan wrote:

>In my experience, here's what happens:
>
>You search the initial position to 10 ply and play e4.
>
>Because of that search, there is a "9 ply" hash table entry for the resulting
>position.
>
>So when you start the next search, you do a 1 ply search(). It finds the hash
>table entry, sees that 9 > 1, and returns immediately. Then you do a 2 ply
>search(), same thing happens. Until you do a 9 ply search.
>
>So it's not necessary to keep track of which entries are still "good" because
>you start the search up again where you left off.
>
>-Tom

Ok, maybe I am just confused here. I thought that if you found an entry in the
"9 ply" portion while searching the 10 ply off the initial position, that entry
may or may not get overwritten due to a hash table collision (hence, the need
for the counter).

So, you will probably not start off where you left off for some nodes as you are
searching them and will start off again for others. Is there some other thing
you do to ensure that a good node in the hash table does not get overwritten
(when doing the 10 ply search)?

KarinsDad :)





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.