Author: Vincent Diepeveen
Date: 11:49:53 09/04/02
Go up one level in this thread
On September 04, 2002 at 14:24:47, Robert Hyatt wrote: With YOUR method it is very easily possible to get always > 2.0. The easy case is a program that gets a speedup of about 1.9 and profits just a bit more than 5% from a filled hashtable. That's by definition > 2.0 then. I need to add that you also have to search in the time of the opponent. Easy scenario. Suppose you take 24 positions. 1 position you get say 1.9 speedup (cleaned hashtable). The other 23 positions you mispredict the move. However the move is a transposition to the search tree the program plays. Then the moves get played and you start a new search with that hashtable. So you get like > 10 ply out of hashtable directly. That saves n minutes calculation where you calculate n minutes 50% of your search is for free then. I don't need to mention speedup is about 2x 1.9 = 3.8 then at 2 processors when compared to someone using a cleaned hashtable each time getting a 1.9 speedup. The average speedup is then : (23 * 3.8) + 1.9 = 3.72 this at 2 processors. >On September 04, 2002 at 14:18:38, Vincent Diepeveen wrote: > >>On September 04, 2002 at 13:36:38, Robert Hyatt wrote: >> >>But you make stupid assumptions. For example, DIEP profits about >>for 40% of the nodes in such a case. Imagine it profits 40%. > >Consider us "even" then. I make "stupid assumptions". You make >"stupid statements"... > > >> >>That's dual with double node count than single. So run A it runs >>with n nodes a second. run B it runs with n/2 nodes a second. >> >>It means a speedup at 2 processors of > 2.0 is very well possible. > >It isn't possible. I have explained why a dozen times. I'm not going >there again. You report > 2.0, and you get lauged at. It's up to you. > > >>It simply says nothing on how a program lineairly scales. the >>'speedup' number is no longer a speedup number but an index which >>is comparable with not a single scientists result. >> >>If your goal is to write somethign that cannot be compared to any other >>research, then i would love to hear it. > >My goal was to answer the question "how much does the parallel search speed >up Cray Blitz in a real game?" > >Nothing more. Nothing less. > >You can ask any question you want. And you can either choose to answer them >yourself, or wait until someone else does. I chose to answer this question >myself, and I did... > > >I don't care about your "hand-waving" about hash table entries. I set up the >best experimental test I could to answer the question. Is it flawed? possibly. >Could it be done better? I doubt it. The very concept of parallel speedup is >non-trivial to measure. Even with fixed-depth searches. A game doesn't have >that luxury so the results are certainly going to be somewhat different. But, >who cares? I want to play games, not run tests...
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.