[petsc-users] SLEPc Eigenvalue problem - Zero Pivot

venkatesh g venkateshgk.j at gmail.com
Thu May 14 12:13:48 CDT 2015


Hi,
1. I am not calling EPSSetFromOptions in source code. I am using the ex7.c
given in the SLEPc webpage.
2. A-matrix is not singular. Only B is singular.
3. I will try with MUMPS.
4. Matrix sizes are same as in Matlab. If B is not singular it is giving
same result as Matlab, but if B is singular it gives the zero pivot issue
which Matlab doesnt give.
So is there a way ?

On Thu, May 14, 2015 at 2:03 PM, Jose E. Roman <jroman at dsic.upv.es> wrote:

> Are you calling EPSSetFromOptions in your source code?
> Is your A-matrix also singular with a nullspace in common with B?
> You can try with MUMPS as explained in section 3.4.1 of the users manual.
> Also check that your matrices are the same as in Matlab.
> Jose
>
>
> El 14/05/2015, a las 08:52, venkatesh g escribió:
>
> > Hi Jose, all,
> >
> > I tried that too, it is telling the same error of "Detected Zero pivot
> in LU factorization".
> > My matrix B in AX=lambda BX has large no. of zeros and is singular.
> >
> > Is there another way ?
> >
> > Venkatesh
> >
> > On Wed, May 13, 2015 at 9:02 PM, Jose E. Roman <jroman at dsic.upv.es>
> wrote:
> >
> > El 13/05/2015, a las 17:28, venkatesh g escribió:
> >
> > > Dear All,
> > >
> > > I am having a block matrix Eigenvalue problem AX = lambda BX
> > >
> > > My B is singular, it is a block matrix and only one block has non-zero
> values.
> > >
> > > Matlab Eigs is giving the result.
> > >
> > > I am using option: "./ex7 -f1 A1 -f2 B1 -st_type sinvert"
> > >
> > > Result is: Detected zero pivot in LU factorization.
> > >
> > > The ex7 is an example from SLEPc.
> > >
> > > Pls advice what to do ?
> > >
> > > cheers,
> > > Venkatesh
> > >
> > >
> >
> > If you know where your wanted eigenvalues are, set the target value, for
> instance:
> > ./ex7 -f1 A1 -f2 B1 -st_type sinvert -eps_target 0.5
> >
> > Jose
> >
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20150514/66149301/attachment.html>


More information about the petsc-users mailing list