Re: A 50%-100% difference in C++ execution time between Solaris and Windows
From: Thomas Dehn (thomas-usenet_at_arcor.de)
Date: 05/26/03
- Next message: Andrew Gabriel: "Re: What is this mean "interrupt level 6 not serviced" ?"
- Previous message: Simon P. Grasser: "php apache build problem"
- In reply to: AviA: "A 50%-100% difference in C++ execution time between Solaris and Windows"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Date: Mon, 26 May 2003 22:33:14 +0200
"AviA" <avia@mercadosw.com> wrote:
> Anybody has any idea why we are experiencing such a difference in a
> c++ code? We are using gcc 3.2, all our code is optimized (-O2), and
> of course the CPUs compared are of the same size. Looks like the
> underlying libstd c++ of gnu or something wrong that we're doing.
For starters, -O2 is a low optimization level only, and
you are using gcc rather than Sun's C++ compiler, which
has better optimization than gcc. Then, if you are making
extensive use of math libraries, you should use
libmopt or similar fast libraries on Solaris. The default libm
on Solaris is slow and exact, whereas the default math
libraries on Windows are fast and inexact.
Thomas
- Next message: Andrew Gabriel: "Re: What is this mean "interrupt level 6 not serviced" ?"
- Previous message: Simon P. Grasser: "php apache build problem"
- In reply to: AviA: "A 50%-100% difference in C++ execution time between Solaris and Windows"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|
