Author: Amir Ban
Date: 16:29:12 09/16/98
Go up one level in this thread
On September 16, 1998 at 18:16:31, John Coffey wrote: >Do programs use hash tables become they spend more time evaluating >the positions at the leaves than they do traversing the tree? > It's not obvious to me that this is relevant. When you have a hash hit, you can get rid of a subtree, which means you neither need to traverse that subtree not evaluate its leaves. Seems to me that whatever the the relative cost in time of these two, you end up saving the same. >How much overhead (CPU wise) do hash tables take? > Should be a few percent. If the overhead is bigger than that, you are better off with no hashing, because you don't stand to gain more than a few percent. It should depend on whether you hash all positions or stop at some depth, and whether you hash into your quiescence depths too. Junior hashes positions down to depth 2, and spends 2-3% on the overhead. >It seems to me that in the endgame you could have the same positions >arise multiple times, but in the middle game it would be rare. Has >anyone done any tests on this? > It's obvious intuitively. I think you'll find some studies in ICCAJ, but I don't have references. Amir
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.