Author: David Rasmussen
Date: 04:37:25 12/04/01
Go up one level in this thread
On December 04, 2001 at 07:06:35, Severi Salminen wrote: >>There is a chance, of course, that it won't affect anything important, but who >>knows? I sure don't want speed for incorrectness. What if this happens on the >>pv? Furthermore, I was under the impression that collisions _shouldn't_ happen, >>even with 32 bits, but it does here. > >Have you thought this: Let's assume that the 32-bit version gives 1 collision in >every 5 seconds. Now, if you use 64-bit you get probably no collisions at all >but you search 1% (or more, or less...) less nodes every second. Which is worse: >searching a node and getting a wrong result or not searching those nodes at >all?? Not searching might also mean "wrong result" in a sense and might result a >wrong PV. And in case of 64-bit keys you will most likely miss more nodes than >the 32-bit version actually collides. > >Just a thought based on speculation. > >Severi Sure, as I said elsewhere, it is a design choice. I personally would give a great deal for correctness, if it can be achieved. That this "problem" is also in Crafty implies that it is not too big of a problem. But still, who knows what impact it has. One could put some debug code into Crafty to check when a collision had a big impact, that is, on the PV, or ultimately, the first move of the PV. /David
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.