Author: Robert Hyatt
Date: 18:39:01 10/08/99
Go up one level in this thread
On October 08, 1999 at 19:40:37, Scott Gasch wrote: >Hi, > >I think my hash table is messed up... I was talking with another programmer >about a test position (Fine 70: 8/k7/3p4/p2P1p2/P2P1P2/8/8/K7 w - - bm Kb1). At >19 ply my program had searched nearly a million tree nodes but his was at like >40k or something. It became apparent that this was because his hash was hitting >like 99%... mine hits about 30%. > >My hash is about 500k positions large and I replace positions if the new one has >an equal or greater depth than the one in the hash already. I've been over my >hashing code carefully and I cannot explain this problem. However I normally do >not see anywhere near even a 30% rate -- it's usually more like 2%. > >My questions are: 1) what kind of hash hit rate (hits / tries * 100) do other >programs see in this position and 2) what could affect a hash in this manner? >Poor key generation is one but I don't think so... too small is another but I >don't think so either... I am at a loss. > >Thanks again, >Scott key distribution is one issue. If your random numbers aren't good, you can end up using only 5% of your hash table or less, and overwriting everything in sight. However, it sounds like something more involved is going on, as in a real bug, as maybe you are not handling the bounds right, or whatever...
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.