Author: Robert Hyatt
Date: 18:37:09 05/02/00
Go up one level in this thread
On May 02, 2000 at 21:00:11, William Bryant wrote: >On May 02, 2000 at 18:55:15, Robert Hyatt wrote: > >[big snip] >> >>simplicity. I want a power of 2 for the size. which means that one has to >>be 1/2 the size of the other, to let the hash table take up 3/4 of memory >>and still be a power of 2. I tried making the depth-prefered table bigger, >>then tried it smaller. Smaller was more efficient, by a fairly small amount. >> >>The alternative would be a combined table with (say) 3 entries per bucket, >>with the first being depth-prefered, the last being always store, and the >>middle one a place to save whatever the other two overwrote. >> > >As I understand it, the idea with the dual entry hash table is that each >hash address holds two hash enteries, 1 depth prefered, 1 always replace. > >Wouldn't this make both tables "the same size"? > >Or are you using two different tables and indexing into each one independently? > >William >wbryant@ix.netcom.com two tables. one of size N, one of size 2*N. The reason is to get around using only 1/2 of memory for the hash table, while keeping a power of 2 size. This allows 3/4 of memory for hashing.
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.