From maintainers-request at octave dot org Mon Jan  9 03:34:30 2006
Subject: Re: Successful compilation with MinGW
From: David Bateman <David dot Bateman at motorola dot com>
To: Ole Jacob Hagen <waterthrill at yahoo dot no>
CC: =?ISO-8859-1?Q?Timoth=E9e_Lecomte?= <timothee dot lecomte at ens dot fr>,         maintainers@octave.org
Date: Mon, 09 Jan 2006 10:29:49 +0100

Ole Jacob Hagen a écrit :

> Hi.
>
> What changes did you make to Makefiles and so on, to compile Octave with
> MinGW/Msys?
>
> How does a MinGW compiled Octave performs versus a Cygwin compiled
> Octave with the same computer? And how does Octave compare with same
> compiler on Linux?
>
> There is a site with benchmark for Octave/Matlab...Don't remember where
> I downloaded it...
>
www.sciviews.org, though this benchmark caused a lot of work in octave 
and so the results presented there are out of date.

> I mangaged also to compile Octave with MinGW (GCC-3.2.3). I've also
> benchmarked these Octave versions: MinGW octave and Cygwin Ocave.
>
> The benchmark results can be found here:
>
> MinGW GCC-3.2.3:
> http://www.octave.org/mailing-lists/octave-maintainers/2005/14
> Cygwin GCC-3.2.3:
> http://www.octave.org/mailing-lists/octave-maintainers/2005/26
>
> Can you do a benchmark test with the new gcc-version?
>
I'm not sure your benchmarks are fair, since you don't install either 
atlas or octave-forge, and I suspect FFTW either, under mingw. This can 
be seen in the relative values you give. The only benchmark value that I 
think is important in these in the Fibonacci test where cygwin wins, as 
this tests for-loops. Though this is with g++ 3.2 and the results would 
have been very different with a later compiler....

Regards
David

> Cheers,
>
> Ole J.
>