Author: Carmelo Calzerano
Date: 15:04:57 02/20/01
Go up one level in this thread
On February 19, 2001 at 14:10:41, Larry Griffiths wrote: >Ok Carmelo, > >I ran Fine70 to a depth of 24 plys. I used two 8Meg hash Tables, One for Black, >One for White. > >00:00:01.984 24 - 1000 Kb1 <---------- > >00:00:01.984 24 - ########## SEARCH TERMINATED (TIME EXPIRED) ########## > >Valid Hash Entries............... 37270 >Refereneced Hash Entries......... 13671 >BTM Hash Entries................. 25036 >WTM Hash Entries................. 12234 > >Elapsed Time= 1.88 seconds > >From my disk log... > >41,829 keys matched at ply 24 and 2,046 keys did not match. >so the hit ratio is around 95%. > >If I divide 13671 by 37270 then only 36.6 % of the hash entries show being >referenced at the end of the search. Well, I think I misunderstood your previous posts. :-) I was speaking about hash key distribution, thinking that your 20-40% was the hash fill value at the end of the search. It seems that (please correct me if I'm wrong) your question is about hash HITS, or better about the percentage of stored hash entries that are reused at least one time later in the search: I never studied that matter in detail, but your numbers looks ok to me at a first glance. :-) Bye, Carmelo
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.