[petsc-users] Setting up MUMPS in PETSc

Jed Brown jedbrown at mcs.anl.gov
Tue Oct 23 18:33:37 CDT 2012


On Tue, Oct 23, 2012 at 6:17 PM, Jinquan Zhong <jzhong at scsolutions.com>wrote:

> Any way to get the condition number.  I used ****
>
> ** **
>
> -ksp_type gmres -ksp_monitor_singular_value -ksp_gmres_restart 1000****
>
> ** **
>
> It didn’t work.
>

First, as a matter of course, remove "It didn't work" from your vocabulary.
Please never utter those words on any mailing list ever again. It wastes
everyone's time.

It is estimating the singular values of the *preconditioned* operator. Use
-pc_type none to remove the preconditioner. Note that this is generally
extremely inaccurate for the smallest eigenvalue, especially for an
ill-conditioned problem and when you don't run to convergence. It can
over-estimate the smallest singular value by many orders of magnitude. In
general, it is difficult to estimate the smallest singular value without a *
solver*. You can use SLEPc to accurately and efficiently compute the
smallest singular values, provided you have a reasonable preconditioner (or
direct solver) for the system.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20121023/6c88533e/attachment.html>


More information about the petsc-users mailing list