Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hashing and draw by repetition

Author: KarinsDad

Date: 21:19:02 08/28/99

Go up one level in this thread


On August 28, 1999 at 16:35:59, Robert Hyatt wrote:

>On August 28, 1999 at 03:06:09, KarinsDad wrote:
>
>
>>Well, it's late and I'm not sure how well I did the math or even if this makes
>>sense. But, let me know what you think.
>>
>>KarinsDad :)
>
>
>OK...now I sort of follow...  you are actually 'building' a tree, much like
>is done in 'breadth-first' or 'best-first' type searches....  Rather than
>doing a traditional alpha/beta search and randomly probe the hash table...
>you are actually walking around the 'hash table' (which really doesn't sound
>like a hash table to me) because it contains the tree you are searching???

Basically. The reason I consider it a hash table is that entries are added to it
and searched within it via a 16K (or 32K, or 64K, depending on what works best)
hash index. Hence, there should be 50 so nodes per hash index on average which
enables a relatively fast search for a given transposition node and hence the
child links are probably not needed (with child links, one dereference per node,
but a 15% increase in node size; without child links, 2 to 8 dereferences per
node if each index has a relatively balanced structure under it).

I will try both and see if the extra speed is worth the extra memory.

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.