Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Somewhat Off-Topic GCC 3.0 compiler results

Author: Dann Corbit

Date: 18:52:36 06/18/01

Go up one level in this thread


On June 18, 2001 at 21:10:33, Pham Minh Tri wrote:

>[snip]
>>Conclusion:
>>For this file (test_fpu.c) Intel is best, followed by MS, followed by either GCC
>>version {seems about a push}.  GCC beats MS at the Crout method, for some reason
>>(which is interesting.)
>>
>>This is a floating point benchmark, and so chess benchmarks may vary markedly
>>from this result.
>>
>>Also, it may be that the Athlon optimizer has a ways to go yet.
>
>You did not say about your option for MS C++ (do you mean MSVC6.0?). Do you set
>maximize speed?

Yes.  Maximize speed, inline any suitable, output for Pentium Pro.

>I believe that MSVC is still one of the best. Last week, I tried to rewrite some
>hot functions by assembly. But one again, I found that MSVC gave me a very near
>perfect assembly code (at least for simple functions). I gave up satisfactorily
>because I could do very little more.
>
>Additionally, the main difference (with me) is price, Intel compiler is so
>expensive, and VC is cheaper and free (under special licence) in my university
>:-)

MS costs us a ton (we have corporate Enterprise Edition) and so the Intel Add-On
(which *requires* MS VC++ anyway) is a marginal extra cost.  For the extras you
get (often superior binaries, vastly superior profiler, etc.) I think it is a
bargain.



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.