Author: Bob Durrett
Date: 16:53:30 12/08/02
Go up one level in this thread
On December 08, 2002 at 18:33:38, Matt Taylor wrote: >On December 08, 2002 at 17:10:58, Edward Seid wrote: > >>The thing I don't like about using Crafty to compare benchmarks is that everyone >>uses different versions of Crafty. In addition, for any particular version, >>there's the Eugene Nalimov compile, the Dann Corbit compile, and who knows what >>other compiles, all giving different bench results. It would be hard to compare >>my machine vs other people's unless we know we're using the exact same version >>and compile. >> >>Perhaps a better choice would be a more established program like Yace 0.99.56 >>which has been the same version for over 1 year. But I don't know if it has a >>bench command. >> >>On December 08, 2002 at 16:31:06, Matt Taylor wrote: >> >>>Sandra is like any other synthetic benchmark: not indicitive of real-world >>>performance. >>> >>>The best benchmark I know of is running an actual chess engine (e.g. Crafty) and >>>reporting nps. > >Well, there are two metrics to consider. You can consider a general optimized >version of a chess engine not optimized for a specific processor. You can also >consider specially optimized forms for each processor. > >In this case, I think the latter is likely more important since people will want >to optimize a particular chess engine for their system. Determining standard >builds is quite difficult, though. Perhaps that's why nobody has done it. :-) > >And yes, any particular chess engine will do. I picked Crafty arbitrarily, but >just about any relatively complex engine will work. > >-Matt As soon as the hardware technology changes, even a tiny bit, your "standard" program will become obsolete. Bob D.
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.