Author: Daniel Clausen
Date: 08:50:52 07/16/04
Go up one level in this thread
On July 16, 2004 at 11:42:44, Eric Oldre wrote: >On July 16, 2004 at 08:59:42, Tord Romstad wrote: > >>Which influence does the branching factor of a two-player, zero-sum game have >>on the efficiency of a parallel search? Is the speedup when using multiple >>processors independent of the branching factor, or does the speedup increase >>(decrease?) when the branching factor increases? Intuitively I would expect >>the speedup to increase with the branching factor, but I have no experience >>with parallel search. >> >>In case it is not clear, when talking about "branching factor" above, I am not >>talking about what is commonly called the "effective branching factor" (which is >>typically in the range 2-3 for a chess program using recursive null move >>pruning), >>but about the average number of legal moves (around 30-40 in a typical chess >>middle game). >> >>Tord > >Tord, >I've heard alot about the top programs having a branching factor of 2-3. But I >don't know how to measure this with my own engine, what statistics do people use >to determine their branching factor. AFAIK, most people just divide the time used for iteration N+1 with time used for iteration N. You can also use visited nodes, but since you most likely measure the time already, it's easier. Sargon
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.