Author: Robert Hyatt
Date: 12:39:03 09/20/00
Go up one level in this thread
On September 20, 2000 at 14:36:13, Mogens Larsen wrote: >On September 20, 2000 at 12:08:31, Robert Hyatt wrote: > >>On September 20, 2000 at 11:52:17, Mogens Larsen wrote: >> >>It is difficult. If there is one positional best move you find near the >>end of the search, do you give the tactical searcher a chance to search it >>for sanity? Suppose that fails? > >If the tactical searcher were to be dependent on the choice of the positionally >best move then it should get a shot at searching the move for sanity. Would it >be possible to, say, force the positional searcher to analyze the second best >move from the initial search if the fail margin is big enough? This would take >extra time of course since that move would have to be cheked tactically as well. > >>I am not quite sure what you mean. In my case, I consider my search to be >>"optimal" in the sense that it is searching the tree that I want it to search. >>Using the current threaded approach simply searches that same tree significantly >>faster. I don't particularly have a requirement that all the CPUs run at the >>same clock speed, although SMP motherboards generally do have that requirement >>to make interrupt delivery 'sane' as well as handling a shared bus. > >I meant simple in the conceptual sense, ie. SMP for the sake of searching deeper >instead of adding complexity to the evaluation. The latter might be the best >chance of advancement in performance for computer programs. The idea of leaving >everything to chip development seem "simple" to me, but I'm not directly >involved so to speak. In that case, don't assume too much. IE I obviously use a parallel search. And it gives me greater depth, which I sometimes choose to trade for more knowledge instead. But nothing says _all_ the extra cpu horsepower is going to just go deeper. This is why I often say that Crafty is not doing well in blitz games, because the assumptions I make about the hardware are wrong for fast time controls on slower computers. however, in allocating multiple processors to some task like chess, it makes more sense (performance-wise) to simply tell them all "go search that tree". IE I would rather have a single large crew building a house rather than several small, specialized crews. Because one of the small crews will often have to wait on others. And while they are waiting, they are not contributing... > >I'll mail you about the ASCII article about SMP in a not too distant future. > >Mogens.
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.