dolfin team mailing list archive

[Johan Hake <hake@xxxxxxxxx>]

On Thursday 08 October 2009 23:28:29 Anders Logg wrote:
> On Thu, Oct 08, 2009 at 11:19:23PM +0200, Johan Hake wrote:
> > On Thursday 08 October 2009 22:39:52 Anders Logg wrote:
> > > On Thu, Oct 08, 2009 at 10:20:24PM +0200, Johan Hake wrote:
> > > > On Friday 09 October 2009 06:35:26 Marie Rognes wrote:
> > > > > Johan Hake wrote:
> > > > > > On Friday 09 October 2009 06:16:59 Marie Rognes wrote:
> > > > > >> Jehanzeb Hameed wrote:
> > > > > >>> Attaching the files.
> > > > > >>
> > > > > >> I get the same result for the two form files with a  DOLFIN >
> > > > > >> 0.9.3:
> > > > > >>
> > > > > >>     Curl error is 223.93
> > > > > >>     Alt  Curl error is 223.93
> > > > > >>
> > > > > >> In other words, try upgrading :)
> > > > > >>
> > > > > >>
> > > > > >> For future reference: Since the issue at hand was comparing the
> > > > > >> two curl implementations, the test could have been further
> > > > > >> simplified by for instance saving the solution u and just
> > > > > >> sending the solution, the two ResCurl*.form files and a
> > > > > >> simplified main.cpp
> > > > > >
> > > > > > While at it, could you take a look at the curl-curl demo too? On
> > > > > > my computer it produces highly different results for cpp and
> > > > > > python. Python takes forever and gives weird results.
> > > > >
> > > > > I'm sitting lazily at Dolfin 7173:fd4071fe9460.
> > > > >
> > > > > The curl-curl demo gives the same result for cpp and python
> > > > > there...
> > > >
> > > > Someone else that have problems with the python curl-curl demo?
> > >
> > > Works very nicely in both C++ and Python for me.
> > >
> > > instant-clean?
> >
> > Have done that...
> >
> > Output from Python (With all JIT calls removed):
> >
> > UnitSphere is experimental. It may be of poor quality mesh
> > Solving linear variational problem
> >   Matrix of size 21424 x 21424 has 872288 nonzero entries.
> >   Applying boundary conditions to linear system.
> >   Solving linear system of size 21424 x 21424 (PETSc LU solver, mumps).
> > Solving linear variational problem
> >   Matrix of size 2187 x 2187 has 81873 nonzero entries.
> >   Solving linear system of size 2187 x 2187 (PETSc LU solver, mumps).
> > Plot active, press 'q' to continue.
> 
> Here's my Python outpout:
> 
> UnitSphere is experimental. It may be of poor quality mesh
> Solving linear variational problem
>   Matrix of size 4184 x 4184 has 61784 nonzero entries.
>   Applying boundary conditions to linear system.
>   Solving linear system of size 4184 x 4184 (PETSc LU solver, mumps).
> Solving linear variational problem
>   Matrix of size 2187 x 2187 has 81873 nonzero entries.
>   Solving linear system of size 2187 x 2187 (PETSc LU solver, mumps).
> Plot active, press 'q' to continue.
> 
> Do you have the latest versions of all packages?

Ahh... I thought so, but an upgrade of FIAT fixed it!

Johan
 
> --
> Anders
> 
> > And output from C++:
> >
> > UnitSphere is experimental. It may be of poor quality mesh
> > Solving linear variational problem
> >   Matrix of size 4184 x 4184 has 61784 nonzero entries.
> >   Applying boundary conditions to linear system.
> >   Solving linear system of size 4184 x 4184 (PETSc LU solver, mumps).
> > Solving linear variational problem
> >   Matrix of size 2187 x 2187 has 81873 nonzero entries.
> >   Solving linear system of size 2187 x 2187 (PETSc LU solver, mumps).
> > Plotting x (unnamed data), press 'q' to continue...
> >
> > As you can see is the assembled matrix ~12 times the size of the matrix
> > in C++. Same size of Mesh, and same element degrees...
> >
> > I have latest everything...
> >
> > Johan
>