Author: Omid David Tabibi
Date: 14:27:38 07/13/04
Go up one level in this thread
On July 13, 2004 at 17:05:11, Dann Corbit wrote: >On July 13, 2004 at 16:41:49, Uri Blass wrote: > >>On July 13, 2004 at 15:37:32, Amir Ban wrote: >> >>>On July 13, 2004 at 13:11:35, Jorge Pichard wrote: >>> >>>>On July 13, 2004 at 11:55:35, Omid David Tabibi wrote: >>>> >>>>>On July 13, 2004 at 11:23:29, Eydun Lamhauge wrote: >>>>> >>>>>>How strong is Ruffian compared to the engines that participated in WCCC 2004? >>>>>>Would Ruffian have made a top three position? >>>>> >>>>>No. Top 5 positions were already taken by those running on quad opterons. >>>>>Assuming Ruffian would have participated with a non-parallel version, it would >>>>>have reached the 6th place at best. >>>> >>>> >>>>To avoid any further disadvantage for the single processor programs, I recommend >>>>two group of winners for next year! >>>> >>>> >>>>Trophies for the SMP programs And for the single processor programs >>>> First group Second group >>>> >>>> 1st Deep Junior 1st Jonny >>>> 2nd Shredder 2nd Falcon >>>> 3rd Diep 3rd IsiChess >>>> >>> >>>While you are at it, why not restrict search to 8 plies ? That would eliminate >>>the disadvantage to FIBChess, for example. >>> >>>Amir >> >>Not enough. >> >>Unfortunately even 8 plies are enough to beat FIBchess easily. >>That program is simply not competitive. >> >>There is also a clear difference between the suggestions and you know it. >> >>single processor winner is at least winner in something with practical meaning >>for people(not everybody is using more than one cpu). >> >>The problem with the suggestion is that deciding about a single processor winner >>in WCCC prevents you to play for both titles so if there is a competition for >>best single processor it should be in different time than WCCC >>and practically there is a competition for best single processor (it is called >>ssdf) and I believe that the ssdf will have no problem to test Jonny or Falcon >>if the programmers ask them to do it. > >I think that all of the top programs will resist a uniform platform event. They >have worked very hard to write SMP programs, and that gives them a huge >advantage. What will motivate them to surrender their advantage? For sure, if >I were them, I would kick and scream and drag my feet and refuse to change the >platform, because it is to their advantage to do so. Not the case with Stefan MK at least. He welcomed the idea of "everyone runs on machines given by organization". > >The choice for the amateur is obvious: >Write a threaded version of your program. It's not that hard. Of course not, and my parallel version was almost ready. But running a parallel engine on a single processor machine is not much fun you know... "Uniform hardware" doesn't necessarily mean single processor machine. If all the participants can be provided with 8 processor machines, that is of course the best option. >Take a look at >Amy's source code, as it is the easiest to understand of the two open source SMP >programs. > >I don't think you will ever see another uniform platform event. Who does it >serve? The amateurs only. The hardware vendors want to sell their top of the >line machines. If they have some 8 CPU whomper, that wins the contest, they can >sell it as "the hardware that won the computer chess world championship" but if >every machine is the exact same machine, there is no way to claim any sort of >superiority. Therefore, they will have little motivation to offer nifty-keen >machines for usage. Similarly, the commercial chess programs have taken a clue >from crafty, realizing that they would get their brains beat in if they did not >write threading programs. The result of this year's contest clearly shows that >the approach was sound. Had all the programs except for Crafty and Diep been >single threading, I think you would have seen a far different result (of course, >then the commercial chess vendors would have been far more amenable to uniform >platform -- wink, wink). > >Multiple CPUs is a very logical step for increased horsepower without stupendous >engineering advances. At some point in the near future, it will be far simpler >to add more CPUs than to increase clock rate (actually, we are already there >IMO).
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.