Any other comments about this?  I'm still thinking about it.

Also... does anyone (Jed?) know how to advance and explicit system alongside an implicit system with TSGL?  What I mean is that I'm solving a transient nonlinear implicit system... but I also have a system of explicit variables that I drag along that also get updated all the time (they are ODE's essentially that I have analytic solutions for).  I already know that if I don't integrate those in time the same way I integrate the implicit system then I get a degradation in time error order.

Any ideas?



On Tue, Jan 26, 2010 at 1:18 PM, Derek Gaston <> wrote:
So... Jed's comments on TSGL today brought to mind something I've been meaning to bring up.

Is there any plan for incorporating time integration packages into libMesh?  We have NonlinearImplicitSystem mapping to SNES or NOX... shouldn't TransientNonlinearImplicitSystem map to TSGL or Rhythmos?

It seems like it wouldn't be much work to adapt the current PetscNonlinearSolver into a PetscTSGLNonlinearSolver.

What do you guys think?