On Wed, 14 May 2008, luyi wrote:

> I change the RKDG code to ExplicitSystem and I write a small Gassian
> elimination solver, it works well and I can get high order results.
>
>
> I have tried the command line "--implicit_neighbor_dofs" and it no 
> improvment.

We may have a bug in sparsity pattern allocation for DG problems,
then.  If your DG example makes the slowdown reproduceable on our
systems, hopefully one of us can find the time to dig into the DofMap
code and figure out the problem.

> Can libmesh run without PETSC but only with mpi?

In theory, yes; that should give you a libMesh DistributedVector
instead of a PetscVector as the default NumericVector implementation.
None of us run that way right now, though, so I can't guarantee there
haven't been any regressions in that code.  There's also no linear
solver implementation (or even sparse matrix implementation, IIRC) for
that case, so you'll have to write your own.  If that works well for
you, I'd encourage you to switch from ExplicitSystem back to
ImplicitSystem, but contribute your matrix/solver code to the library.
PETSc has a thousand features and a million options, and we should be
getting Trilinos support sooner or later, but having a lightweight
solver internal to the library would be nice too.

> I'd like to send a DG example after several days, libmesh is a so excellent 
> library and thank you for your kindly reply.

Thanks!
---
Roy

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft 
Defy all challenges. Microsoft(R) Visual Studio 2008. 
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to