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

[Dominik Szczerba]

> What makes you think that the point where some nonlinear solve stagnates has> anything to do with correct time-averaged statistics?
> What is your verification and validation process?

Strictly mathematically you are probably right. Intuitively
physically,  I do expect - and in fact get - some representative
solution.
I validated only qualitatively, looking at streamlines. I do get the
expected vortices at the expected locations by eye looking very
similar to the ones in published papers.
> You should use a turbulence model. Even a zero-equation model like
> Smagorinsky would be better than simple streamline diffusion.

You are right, but I will not be able to implement it in the nearest
future. And the focus is transient von Karman problem, steady state
was just a side jump to see what it would do. Streamline diffusion
just proved your guess right about the singular system. Still I am not
happy it crashes instead of diverging.

Thanks
Dominik