[petsc-users] Fieldsplit with LSC for constrained elasticity/poroelasticity?

[Matthew Knepley]

On Tue, Oct 28, 2014 at 9:03 AM, Jed Brown <jed at jedbrown.org> wrote:

> Matthew Knepley <knepley at gmail.com> writes:
>
> > On Tue, Oct 28, 2014 at 8:00 AM, Tabrez Ali <stali at geology.wisc.edu>
> wrote:
> >
> >>  Matt
> >>
> >> With MUMPS it is indeed the same but not faster. E.g., see 1.txt and
> 2.txt
> >>
> >
> > There is a big problem here. This system is never actually being solved.
> > Look at 1.txt:
> >
> > 0 KSP preconditioned resid norm 9.320055451716e+05 true resid norm
> > 1.755998647494e+02 ||r(i)||/||b|| 1.000000000000e+00
> > 380 KSP preconditioned resid norm 9.034871458425e-05 true resid norm
> > 3.118451181896e+09 ||r(i)||/||b|| 1.775884728811e+07
> >
> > The preconditioned residual did drop 10 order of magnitude, but the true
> > residual (b - Ax) is still very very large.
> > This looks like the system is singular and the preconditioner is masking
> > this. This needs to be figured out first.
>
> It could also be very poor scaling between the fields.  It's important
> to nondimensionalize so that you don't ask to solve for velocity 1e15
> more accurately than pressure, or vice-versa.
>

Yes, this was a big deal in PyLith. There is a large gap between
displacements and
the Lagrange multipliers. Without non-dimensionalization, we could not
solve the system.

   Matt

-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20141028/dac390cc/attachment.html>