Re: A plea for gsvd

[David Bateman]

Pascal Dupuis wrote:

>On Tue, Apr 11, 2006 at 06:55:12PM +0200, David Bateman wrote:
>
>  
>
>>This then comes down to two questions. 1) Who is going to do it, and 2)
>>How much are you willing to pay them. 
>>    
>>
>
>OK. I had time to rant, so I did a first try. Enclosed is a tar file
>containing : 
>liboctave/gsvd.cc : generic interface, modelled after svd.cc
>src/DLD-Functions/{Cmplx,dble}GSVD.{h, cc}
>plus modified Makefile.in to add those 'features'
>
>I'm not a C++ guru, so I took the corresponding files from *SVD* and
>made something similar, no inheritance whatsoever.
>
>It was developped inside octave-2.1.73, with debian patches, and
>linked against lapack3. I guess this should not change the final
>result. 
>
>The problem is that a call to gsvd results in a segmentation
>fault. I probably made some basic mistake with the function call,
>could someone please check the source for some obvious error ?
>
>TIA
>
>Pascal Dupuis
>
>  
>
Pascal,

Given code as a basis I'm of course willing to help out, and give
suggestions. Firstly, I think as John says, if we are going to modify
the SVD classes it must be done against 2.9. However, I'm not sure if it
is better to derive a new GSVD class from SVD or where it is better to
treat both svd and gsvd in a single class.

Give me a few days and I get back to you, I'm a bit busy now...

D.

-- 
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

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