On Thu, Oct 23, 2008 at 4:38 PM, Derek Gaston <[EMAIL PROTECTED]> wrote:
> Ok - I wasn't quite enamored with John's last diagram... so I've attached
> my own. In particular the multiple inheritance from UserShellMatrixBase to
> the Petsc / Epetra MatrixBases made me feel icky.
>
> The idea is similar to the structure we have for NonlinearSolver... where
> there is a PetscNonlinearSolver and a NoxNonlinearSolver... but you just
> attach a residual() or jacobian() function to whichever nonlinear solver
> gets built with build().... not caring if it's from Trilinos or Petsc.
>
> I tweaked this idea slightly though... by attaching a class instead
> (UserShellMatrix).... just to encapsulate things a little bit. We could,
> instead, just attach functions with a specific signature that implement
> things like mult()... but I think it will be nicer to have UserShellMatrix
> objects. Either way it's the same thing.
>
> Now... if you want to add a PetscMatrixBase and an EpetraMatrixBase to
> this... you can do that... and just have PetscMatrix and PetscShellMatrix
> inherit from it.
>
> The nice thing about this design is that user code doesn't need to know
> which solver package you are using... at the same time a ShellMatrix can be
> passed into routines that require a NumericMatrix... etc.
Interesting...
We want ShellMatrix to be usable by either Petsc or Epetra, so I don't see
the reason for creating separate PetscShellMatrix and EpetraShellMatrix
subclasses...
--
John
-------------------------------------------------------------------------
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
