Author: Robert Hyatt
Date: 16:01:22 09/04/02
Go up one level in this thread
On September 04, 2002 at 17:48:19, Gian-Carlo Pascutto wrote: >On September 04, 2002 at 15:01:55, Robert Hyatt wrote: > >>On September 04, 2002 at 14:49:53, Vincent Diepeveen wrote: >> >>>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. >> >>No, that is by hand-waving > 2.0. Because the serial search would >>_also_ profit by the filled hash table and run faster. >> >>Try again... > >The problem is that one position influences the other. They are >not independent measurements, but your article is based on the fact >they are. > >-- >GCP And does it not say that? Wasn't the _base premis_ to measure the speedup during a _game_??? If you can define a better methodology that I did, by all means do so and I'll try to use that next time. However, the issue was speedup in a game, and in a game _all_ the positions, and as a result all the speedups, _are_ related... And I do _not_ see any problem with that, from any scientific perspective I can think of. I asked a question. I formulated an experimental setup to answer the question, I ran the experiment, and reported the results... If Vincent doesn't like the "game" measurement, he can use my PhD dissertation data instead, which was 5 ply searches on 1 to 16 cpus... I think the peak speedup was 9.0 on shallow searches on unrelated positions....
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.