[petsc-dev] SNES for 2x2 problems

[Jed Brown]

On Wed, Jan 2, 2013 at 4:33 PM, Matthew Knepley <knepley at gmail.com> wrote:

> And if it does not converge with fixed step? It will eventually morph nto
> SNES.
>

You have a _serious_ problem with your mesh if a simple method does not
converge.


> I think your approach is bogus unless you can prove that your simple thing
> is sufficient.
>

Tons of low-dimensional rootfinding problems never require the full
machinery. I'm asserting that this is one of them based on the fact that
every library I've seen has a very simple implementation that doesn't cause
any problems.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130102/3e940579/attachment.html>