[petsc-users] reg: Eigenvectors of general complex eigen value problem

venkatesh g venkateshgk.j at gmail.com
Mon Mar 24 04:28:37 CDT 2014

Ok. I am doing the following

1. Running "./ex7 -f1 A2 -f2 B2 -st_type sinvert -evecs VC"
2. In MATLAB I do "vc=PetscBinaryRead('VC');"

The output is one Eigenvector in VC which is 800x1.

And I did the normalization vc=vc/max(vc) and compared with the normalized
original vector.. Also I must tell that if A2 and B2 are real, it works.

I also changed the line 179 of ex7.c like u said.

Pls let me know.

On Mon, Mar 24, 2014 at 2:24 PM, Jose E. Roman <jroman at dsic.upv.es> wrote:

> El 24/03/2014, a las 05:59, venkatesh g escribió:
> > Ok. I used -st_type sinvert only,
> >
> > I am comparing it eigs after normalization. Even now the eigenvectors
> differ.
> >
> > Whether I should use a different eps solver ?
> >
> > Pls let me know.
> >
> It works for me, so you should provide more details about how you are
> loading the eigenvectors in Matlab and comparing them.
> Note that the complex version of the example also writes the xi vector
> (which should not) so you will have zero vectors interleaved with the
> eigenvectors. If you want to avoid this, make the following change (line
> 179 of ex7.c):
> #if !defined(PETSC_USE_COMPLEX)
>        if (!ishermitian) { ierr = VecView(xi,viewer);CHKERRQ(ierr); }
> #endif
> Jose
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20140324/4257d2c0/attachment.html>

More information about the petsc-users mailing list