Author: Eugene Nalimov
Date: 22:19:22 01/10/01
Go up one level in this thread
On January 10, 2001 at 16:25:59, Dann Corbit wrote: >On January 10, 2001 at 16:13:06, Robert Hyatt wrote: > >>On January 10, 2001 at 15:37:40, Dann Corbit wrote: >> >>>Here she be: >>>ftp://cap.connx.com/pub/chess-engines/new-approach/CRAFTY1.EXE >>> >>>SMP Version here: >>>ftp://cap.connx.com/pub/chess-engines/new-approach/CRAFTYSMP.EXE >> >>OK... I give up... what is "new approach"? >> >>:) > >Not much to it. > >I change a few compiler settings, and use the Intel C++ 5.0 compiler do most of >the work. She's a real honey. > >I let the compiler agressively inline and target P6 and above chips. I also >change the function calling convention and a few other odds and ends (you've >seen my dirty #include trick, for example). > >As far as making a fast binary, the Intel C++ compiler is the best there is. > >If someone else wants to try it, you should buy both versions 4.5 and version 5, >since there are some files that won't compile because it is an incredibly picky >compiler. I use this procedure: >Compile with Intel C++ 5.0 >(for all modules left not compiled): >Compile with Intel C++ 4.5 >(for all modules left not compiled): >Compile with Microsoft C++ 6.0 (with processor pack installed). > >Then, for obvious gaffes, I repair them one at a time and repeat the process. >That way, I have something that runs right off the bat, and eventually, a real >bolt of lightning. So, generally you are saying "Intel compiler does not work". AFAIK, main objective for Intel compiler is to produce good results on SPEC benchmarks. What will happen with real-world code does not matter. Eugene
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.