Author: Peter Fendrich
Date: 05:16:56 01/10/03
Go up one level in this thread
On January 10, 2003 at 08:01:45, Dan Andersson wrote: >A couple of postings on comp.lang.scheme implemented the CoyoteGulch benchmarks >in Scheme. And lo and behold BigLoo produced faster code for the numerical >benchmark. One might object that since BigLoo emits C or Java bytecode it isn't >really faster. But the amount of automated program transformations that are >applied is huge. For a coder to do the same thing would be like trying to >outperform a spreadsheet. And the C code is inhuman in nature. And the question >arises: Why on earth would one use C++ or Java? Both are verbose and terribly >low level compared to lambda calculus. > >MvH Dan Andersson For number crunching yes, but chess programmimg is a lot more than number crunching. You are dealing with complex data structures that are moved around and tranformed back and forth. We also have all the chess knowledge intensive code with such a complex logic that "global" optimizing is not efficient. I have seen benchmarking beteween Ocaml and C where Ocaml is more efficient than C for a wider range of tasks than only number crunching. It is all meassure with many, but rather small isolated functions. Put it all together in one complex system and the ocaml optimizer wont find it's way out anymore. I haven't tested it though... Mvh(!) Peter
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.