Author: fca
Date: 12:06:59 08/12/98
Go up one level in this thread
On August 12, 1998 at 14:26:10, Mark Young wrote: >On August 12, 1998 at 14:16:03, fca wrote: > >>On August 12, 1998 at 11:46:18, Mark Young wrote: >> >>>On August 12, 1998 at 11:13:06, fca wrote: >>> >>>>If multitasking two chess programs on one PC, and playing a game between >>>>them, how does one "fairly" allocate resources between them? >>>> >>>>I accept there is no question of properly deducing playing strength in such a >>>>scenario. That is NOT the objective, therefore. >>>> >>>>0. What does "fairly" mean? >>>> >>>>1. RAM? >>>> >>>>2. CPU %? >>>> >>>>3. Idle / background priority? >>>> >>>>4, Thinking Time set on program? >>>> >>>>and especially >>>> >>>>5. Should permanent brain be switched off on one or both programs? >>>> >>>>Ideas on how to arrive at the answers, please, or better still - the answers >>>>themselves. :-) >> >>>When I test programs on one computer I used a CPU meter, so I would know if one >>>of the programs was taking any cpu. Some programs will, even with permanent >>>brain off. >> >>Programmer guesses a competitor may be running... So keeps reading the keyboard >>intensively, or doing some other hogging. They are a sly lot. Of course >>motives cannot be attributed for sure! ;-) >> >>> Or some programs like Nimzo98 do not have permanent brain option. In >>>either case what you need to do is open up an option box in the program like >>>level or database or something like that. Check the cpu meter till you find >>>something that stops the program for taking any cpu.Then have the other program >>>think while the option box is still opened in the other. >> >>Good suggestion! >> >>BTW, I suspect (by virtue of solving time variation, not attributable to hash >>table effects as program probably had none and anyway the behaviour was >>reproducible from clean start-up) one program of starting thinking even in setup >>position, as you enter the pieces, once kings are on the board. It was for the >>ZX Spectrum, so I cannot check any more. It either assumed WTM or needed that >>input of Who to Move before commencing Setup! Of course, each time you >>added/subtracted a piece, it reinitialised thinking. Cunning way of improving >>solving times. >> >>Re all the other points, Mark? >> >>Kind regards >> >>fca > >All the other points are pretty much common sense. I disagree. I think there are a number of pitfalls. "common sense" is one of the phrases which produce Pavlovian responses (so I'll skip the 'uncommon') ;-) >If you are testing the programs keep things a equal as possible. No, as said this is nicht for testing - I quote: >>>On August 12, 1998 at 11:13:06, fca wrote: >>>>I accept there is no question of properly deducing playing strength >>>> in such a scenario. That is NOT the objective, therefore. Mr N may well have two machines available - but they are liable to be of such differing strength (we keep hearing of these 386s) that this might *actually* be an option which gives him more enjoyable chess. > Time, He may want to "even" the programs up in some way by giving one more time than the other. And he may have no way of switching permanent brain off... I have asked in my reply to Bob's post for a calculation of the performance effect of different time settings on programs co-running with permanent brain on. It is a nasty computation. I have done a draft, but it is more valid to have someone else do it independently and publicly compare it than to post your own and get it glanced at rather than checked. Please take a punt at the problem. > ram Hmmm... Is equality the same as equity? And is it making best use of the resources?? Say, I have an average machine (nowadays here, that probably means something like a P2/233 with 32Mb). I want to co-run Hiarcs and Fritz at 40/2 - faster time controls will exaggerate the effect of my human "move transfer" times. Now at 3 mins/move, even getting 50% of cpu time, Fritz can fill up oodles of hash while Hiarcs won't. So giving them 14Mb each or whatever is not getting the best out of the system. Why handicap Fritz if there is no benefit to Hiarcs. Better chess was an implied objective... >, ect. The hard part is >stopping the other program for taking CPU time when its not its turn to move. But you may want it to, because of the unfair penalty on the better-guesser. You may even want to differentially allocate times to compensate.... Please - there is much here, it is not a trivial matter IMO. (Systematically approaching it of course cannot be done in this medium, I know). I am sure you can have much more gainful input. Kind regards fca
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.