Re: fft

[David Bateman]

According to Michael Creel <address@hidden> (on 04/23/04):
> Hi, I saw your message to Robert French about fft. I am running octave 2.1.57 
> as obtained from debian linux (unstable branch), on the i386 architecture.
> 
> ||/ Name               Version            Description
> +++-==================-==================-====================================================
> ii  octave             2.1.57-2           The GNU Octave language 
> fornumerical 
> computations (
> 
> 
> I have fftw 3 installed:
> 
> ||/ Name               Version            Description
> +++-==================-==================-====================================================
> ii  fftw3              3.0.1-9            Library for computing Fast Fourier 
> Transforms

You might have fftw3 installed but that is no guarentee that octave uses it,
as the decision is made at octave build-time. Therefore the only guarantee
is to check octave_config_info for FFTW3...

> The results I get with your examples are:
> x = randn(300,2); y = fft(x,512);  -------> OK
> x = randn(600,2); y = fft(x,512); ----------> segfault
> x = randn(300,2,2); y = fft(x,512); -------> OK
> x = randn(600,2,2); y = fft(x,512);----------> segfault
> x = randn(2,300,2); y = fft(x,512,2); ----------> OK
> x = randn(2,600,2); y = fft(x,512,2); ----------> segfault
> 
> Hope that is useful, if you would like more information just ask. Michael

Ok, the segfault happens when the size of the matrix is shrunk prior to
the FFT being performed. So I've figured out why I don't see the problem.
There were, some recent changes in the resize_and_fill and resize_no_fill
functions in Array.cc that addressed some bugs in these function. As I'm
using the CVS version of octave, I therefore have this fix installed and
don't see the problem. However, I went back and tested with a version of
2.1.56 I had installed and I now get the same behaviour.

So, depending on how pressed you are to have a solution, you can wait for
a 2.1.58 release, and its debian build, or go and build the octave
CVS yourself..

Cheers
David


-- 
David Bateman                                address@hidden
Motorola CRM                                 +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 1 69 35 77 01 (Fax) 
91193 Gif-Sur-Yvette FRANCE

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------