Author: Robert Hyatt
Date: 19:58:28 08/21/02
Go up one level in this thread
On August 21, 2002 at 18:31:05, Gian-Carlo Pascutto wrote: >On August 21, 2002 at 17:55:44, Robert Hyatt wrote: > >>It doesn't change _that_ much on the last 3-4-5-6 searches. At least it >>doesn't for me in the results from Cray Blitz on 16 and 32 cpus... > >32 on shared memory is a far cry from 480 in a 3 layer system Not particularly. The penalty is in extra nodes searched, more than anything else. Or in the case of Deep Blue, the time lost because the SP2 couldn't keep the chess processors busy 100% of the time, as has been explained previously... Searching in parallel is easy. Avoiding extra work is hard. > >>Once >>those two programs get to the 1-2 second mark, the cpus stay busy, period. >>Just like Crafty. I don't believe DB had any such problem either, once >>you get beyond the first search result they outputted for a iteration... >> >>You could test your hypothesis however, by seeing if the first couple >>of branching factors are worse than the last couple for each search... >>If they are similar, this isn't a problem... >> >>I might try this tonight late, for one log file... > >I doesn't work. You have two unknowns (NPS and BF) and one equation (the data). > >You can't infer both. > As a general rule, NPS tracks time pretty well. Within enough reason to make a conclusion. If _every_ search they do hits around 4.0, it suggests that they are _not_ getting more efficient as the search gets longer. Which is not terribly surprising, once you get beyond the 1 second total searches... >-- >GCP
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.