Thanks Luca and Wolfgang. The implementation is much clearer now. I
will try coding a model evaluator and see how things proceed.
--Shiva
Quoting Wolfgang Bangerth <[email protected]>:
Krishna,
I believe deal.ii doesn't have wrappers for these two specific
packages. Is their any alternative?.
There aren't any wrappers, but all TrilinosWrappers classes have a method to
get a reference or pointer to the underlying Epetra matrix or vector class,
which you could then feed to NOX and TEKO.
If their is no alternative, I am considering writing my own wrapper to
these packages (on the lines of deal.ii supplied Trilinos wrappers),
and link the additional Trilinos libraries by modifying the default
deal.ii makefile linking options. Are their any deal.ii specific
potential compilation/linking issues I should be aware of?.
The only thing I see is that one needs to modify the deal.II Makefile. The
list of libraries one needs to link with is available from one of the CMake
files that Trilinos copies into the directory into which Trilinos is
installed, but we don't currently use this information. That's definitely
something that would be worthwhile doing, since it makes the deal.II Makefile
work with any set of Trilinos libraries people may want to compile for.
As for wrappers: I'm slightly ambivalent. There are cases where that's
definitely useful, for example for linear solvers or preconditioners, because
there are existing interfaces in deal.II with similar functionality and where
it would be nice to provide Trilinos functionality that follows the same
interface.
On the other hand, it may not be worth it wrapping all the interfaces that
Trilinos has through deal.II, primarily because no similar functionality
exists in deal.II and consequently there is no need for interface
compatibility or similarity. Depending on what functionality you
want to wrap,
there may or may not be any other possibly use of these interfaces within
deal.II. You may want to think about whether it is useful in your case, or
whether it wouldn't be simpler to just extract the Epetra object from the
wrapper classes and then code directly against NOX/TEKO.
Best
W.
-------------------------------------------------------------------------
Wolfgang Bangerth email: [email protected]
www: http://www.math.tamu.edu/~bangerth/
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
