[petsc-users] Infinite or not-a-number generated in norm!

[Jed Brown]

On Fri, Dec 30, 2011 at 10:59, Dominik Szczerba <dominik at itis.ethz.ch>wrote:

> Yes, I am a bit malicious, solving for steady state where such is
> known not to exists (Karman vortex street past a cylinder). The very
> same problem (mesh, constraints, etc.) does not crash when solving
> with time stepping. However, I am not completely sure if this is the
> reason of the crash, because I am expecting to never be able to
> converge, but not to crash.
>

You could very well be producing a singular linear system (with large
unknown null space). I'm not sure why you want to do this. There are
methods to solve for periodic behavior and you could choose a turbulence
model in order to produce a steady state.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20111230/09d54035/attachment.htm>