Re: [Help-gsl] Eigen solution of non-symmetric tridiagonal

[Patrick Alken]

Yes, LAPACK routines for eigenvalues/eigenvectors are much more advanced 
than the gsl ones, especially for the non-symmetric cases.

For your particular problem, you would need to use the gsl_eigen_nonsymm 
routines which handle a general non-symmetric matrix. There are no 
routines in GSL which specifically handle nonsymmetric tridiagonal 
matrices. There may be such a routine in LAPACK.

Patrick

On 03/24/2011 09:24 AM, John Chludzinski wrote:
> First, I would consider using LAPACK. I use LAPACK with Cygwin (it one of
> the math packages that can be downloaded with Cygwin).
>
> My experience using GSL to extract eigenvalues and vectors has been painful.
> I used GSL to do a Cholesky factorization, Householder
> similarity transformation,and finally solve for eigenvalues and vectors on a
> tridiagonal matrix. On a 4000x4000 matrix the cumulatively time was ~9 hrs.
>   On the same (AMD) machine using LAPACK, it took ~3.5 minutes.
>
> ---John
>
> 2011/3/19 Dr. Mehmet SAHIN<address@hidden>
>
> > Dear Sir/Madam,
> > We try to find eigenvalues and eigenvectors of a non-symmetric tridiagonal
> > matrix. But we have not found any solution in the GSL subroutines. How can
> > we use the GSL subroutines to solve eigenvalues of a nonsymmetric
> > tridiagonal marix?
> > Our real matrix T has the property that the products of pairs of
> > offdiagonals T(i,i+1)*T(i+1,i) are all positive. Which subroutine(s) we
> > should use? Could you help us about solving this problem?
> >
> > Sincerely,
> >
> > Mehmet SAHIN
> > Selcuk University
> > Department of Physics
> > Konya, Turkey
> >
> > e-mail: address@hidden
> >
> > _______________________________________________
> > Help-gsl mailing list
> > address@hidden
> > http://lists.gnu.org/mailman/listinfo/help-gsl
> _______________________________________________
> Help-gsl mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/help-gsl