Author: Bruce Moreland
Date: 10:32:54 12/27/02
Go up one level in this thread
On December 27, 2002 at 04:54:42, Dan Andersson wrote: >What Bruce is refering to is the idea that if a hash collision happens you >simply put the offending hash in the next available slot. The possible problem >is that you get a chaining. I.e. a long sequence of occupied slots. But that is >rare. I prefer using a recoding algorithm that computes a new index, the net >result is the same. The hash should be at least twice as large as the depth of >search to get good behaviour. A mean number of probes less than two. One other >idea is to create a new hash table each time a capture is made. The question is >if the extra bookkeeping will make it a win, I'd guess no. I don't care if I am get chaining, because the observed performance is good. If the table has a thousand elements, the odds of even getting a collision when you hash a few dozen elements aren't that high. bruce
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.