On Mon, 15 Sep 2008, Derek Gaston wrote:

> This is really great stuff.

You've just realized that you could have entirely avoided duct taping
that mesh smoothing code to ours if you'd only been willing to do this
plus a finite differenced residual, huh?  ;-)

Oops, that reminds me: the finite differenced Jacobian capability
currently won't give correct matrix entries for tightly coupled moving
mesh problems... I'll fix that shortly.

> We're still not using FEMSystem (maybe someday I'll merge my stuff
> with yours...), but it sounds like we'll be able to do this external
> to FEMSystem as well... right?

It depends on how tightly coupled you need to be, I think.  It would
be possible to do an FEMSystem for the mesh movement and a
non-FEMSystem for the physics, for example, but only if you do some
kind of operator splitting in the time integration.  The nice thing
about doing it in FEMSystem is the TimeSolver classes then know about
the mesh movement and can, for example, move the mesh points to the
three different locations you'd want to evaluate pre-timestep,
post-timestep, and mid-timestep for Crank-Nicholson.

> As long as you're adding stuff that people have been looking for for a 
> while... how about restricting variables to subdomains? ;-)

We'll see.  The moving mesh stuff was necessary for this ablation-only
test application, which I needed to code up from scratch, but it looks
like at least for now our complete problem is going to be solved via
that "duct tape" strategy.  So for me to do major libMesh work, I'd
either need enough clout to pass DPLR work off to others, enough
Fortran skill to do all the code coupling and UQ work with plenty of
time to spare, or some sort of supernatural intervention.  I've got my
fingers crossed for leprechauns.
---
Roy

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Libmesh-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to