Re: Differing results on right division

[Damian Philipp]

Hello,

Am 27.05.2011 15:22, schrieb Jordi Gutiérrez Hermoso:
> On 27 May 2011 07:09, Damian Philipp
> <address@hidden> wrote:
> In case anyone else besides me finds rapidshare annoying, I've
> re-uploaded the file here:
> 
>      http://jordi.platinum.linux.pl/accuracy-demo.tar.gz

I too find rapidshare quite annoying and only chose it for a lack of
alternatives, so thank you for rehosting this.

Over the weekend I fiddled around a bit and can now report that this
issue still appears when compiling both my (demo-)program and octave
3.2.4 using the latest library versions (IT++ 4.2.0, ATLAS 3.8.4, LAPACK
3.3.1, all built on Debian Lenny using GCC 4.3.2 (gcc, g++, gfortran)).

[Side Note: I am not entirely sure whether octave is using the correct
version of ATLAS. ldd tells me that my octave binary links to my
compiled version of libatlas as well as the system-provided libatlas,
probably due to many other (system-provided) libraries such as fftw3
being included in Octave, which in turn reference the system-provided
libatlas. Nevertheless, in Octave, the calculation appears to work
correctly whereas it does not in my program.]

Regards,
Damian Philipp


-- 
 University of Stuttgart, IPVS    phone: +49 711 685-88339
 Distributed Systems Group        www.ipvs.uni-stuttgart.de
 Universitaetsstrasse 38          address@hidden
 D-70569 Stuttgart, Germany