Hi Derek,

Derek Gaston wrote:
> All Trilinos support is still very much in development at this point. 
> If you look closely you'll see that it is very incomplete and what is in
> there is just hacked up at this point.  In particular, the NOX stuff is
> currently limited to solving Matrix Free at this point... with no

Ah, speaking of which: As far as I can tell the only missing parts for
providing an explicit Jacobian are the EpetraMatrix.swap() (is this
really needed for Trilinos?) and the EpetraMatrix (Epetra_FECrsMatrix
&m) constructor. I saw you commented out those two funtions in the
EpetraMatrix.h/C files. Do you remember why? I mean, I did less than
thorough testing, but the explicit Jacobian variant seems to work fine
with these lines uncommented, at least on a single CPU. In parallel, the
same bug as with marix-free solving occurs.

> capability for any preconditioning.  Further, Aztec stuff hasn't been
> properly hooked up to accept the new "tolerance" scheme, so there is no
> way to change convergence tolerances.
>
> These are just a coupled of the shortcomings... the point is... it's not
> done.
> 
> Now... with that said... if you want to pitch in I would really
> appreciate the help!

I understand that Trilinos support is unofficial and work in progress. I
am still new to libmesh, and Trilinos in particular, but will invest
some more hours into this problem. Not sure if that will be of any help,
though...
What puzzles me is that the error is already fully present after the
first nonlinear iteration, ie. after solving the linearized system.

> I don't currently have a good build of libmesh with trilinos to do any
> testing on the Laplace Young problem with NOX... but I'll be getting
> back to that soon.  I was pretty sure though that everything was working
> in parallel when I last committed stuff... so I can't say off hand why
> it wouldn't work.  I did a lot of work to make EpetraVector work
> properly in parallel...

Yeah, I got that impression from looking at the devel mailing list ;-)

Norbert

PS: And sorry for the double post. The university here seems to mess
things up with their own reply-to headers...

-------------------------------------------------------------------------
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-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to