Author: Sune Fischer
Date: 02:12:28 06/21/02
Go up one level in this thread
On June 21, 2002 at 04:01:41, Sven Reichard wrote: >Maybe I'm missing something here, but if you furthermore agree to use a hash >table of size 2^i, i <= 32, you can replace all mods, divs and muls by a 32 bit >'and'. The performance hit of the slightly smaller table (less than factor 2) >should be outweighed by the faster access. > >just my 2 bits How do you know that? If you use power 2 size then you get pretty big jumps; 64,128,256,512. So you really have only 4 sizes of the hash. I think it is better to use 200 MB with modulo than 128 MB with AND (there are not divs or muls), the save is not _that_ much compared to what you save when getting those extra hashhits. Of course I don't _know_, I'm just guessing :) Some claim the distribution of the keys are better if the number of entries is prime, though I'm not convinced of that I can't rule it out. just my two øre. -S. >Sven.
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.