Re: [Lapackpp-devel] Computing a single eigenvector

[Christian S. <sti...@tu...>]

Hi!

Am Donnerstag, 3. April 2008 20:14 schrieb Cody Planteen:
> I am using LAPACK++ with some very large symmetric matricies for which
> I need to compute eigenvectors.  However, I only need the eigenvector
> associated with the second smallest eigenvalue, rather than all
> eigenvectors of the matrix.  Are there any LAPACK++ functions to
> compute a single eigenvector? 

No.

> I have done some research on LAPACK 
> functions and the "expert" function SSYEVX allows for computation of a
> range of indicies, rather than all eigenvalues/eigenvectors.

If SSYEVX does what you want, you should probably use it directly in *your* 
source code, rather than trying to fit it somewhere into LAPACK++. I think 
this would be the easiest solution.

> If I 
> were to modify the LAPACK++ source to implement this, are there any
> naming conventions for the function or other considerations that I
> should be aware of?

There are almost no naming conventions. However, I doubt whether there is an 
easy possibility to design a C++ function that offers all functionality of 
SSYEVX through some meaningful arguments, hence I'd rather propose you should 
use SSYEVX yourself. But if you think it's easier for you to do this in 
LAPACK++, we're open for patches :-)

Christian