Author: Dann Corbit
Date: 15:41:37 07/03/01
Go up one level in this thread
On July 03, 2001 at 18:23:53, Carlos del Cacho wrote: >On July 03, 2001 at 16:59:17, Dann Corbit wrote: > >>On July 02, 2001 at 18:58:07, Carlos del Cacho wrote: >>[snip] >>>Anyway there are dubious claims in that document, I don't see how you can make a >>>move generator 30 times faster than crafty's, so I doubt if he ever got to >>>implement it, and if he did what was the HW difference he used to measure it. >>>This person sounds like a day dreamer to me. >> >>I suspect his evaluation will do nothing at all besides material. >>At any rate his move gerator target is 10M NPS. >>Crafty's move generator is 2M NPS. That's only 5 times faster. > >I was just quoting the author here. Five times is feasible but I don't think >this is possible: > >" In assembly language I have succeeded in writing a legal move generator that >is THIRTY TIMES FASTER than Crafty!! " > > >>And he plans to use DDR memory, which is twice as fast. That means 3x faster >>should do it. > >That's not a merit of the software. I can plan to use a 1 GHz CPU and compare >the results with Crafty running in a plain Pentium Pro. A little youthful exuberance isn't so bad. I think there are a lot of mistakes on his page, but I also think despite his ego and incredible... *hyperbole* he's a pretty smart kid. I think his move generator might be a valuable asset if he ever gets it perfected. He certainly knows a lot more about chess programming than I did when I was 16. >>I would not be surprised if it was capable of that. Doesn't mean it will play >>good chess though. >> >>A good move generator is a necessary condition for a world-beater program. But >>it's nowhere near sufficient. > >I'd say all you need is a move generator. There are things that are much more >important and you can always go back to do the tweaking when you get something >that plays fine. And overoptimizing it is just a waste of time. Premature optimization is a tremendous source of evil. 1st rule of optimization: "Don't do it." 2nd rule of optimization (for experts only): "Don't do it yet." Maybe you can get by with a slow move generator. I have never seen a program that plays excellent chess with an extremely slow move generator. And I have seen programs with a decent move generator where a profiler reveals it is a bottleneck. But all in all, the search is much more important than the move generator. A 30x faster move generator with negamax and lame move ordering will get toasted by a good negascout algorithm with null-move and good move ordering. And the quality of the evaluation is probably more important than the speed of the move generator[1]. [1] When asked, "How many moves do you see ahead?", Capablanca said: "One move - the best one." I wish I had his move generator. ;-)
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.