dolfin team mailing list archive

["Garth N. Wells" <g.n.wells@xxxxxxxxxx>]

Anders Logg wrote:
> On Mon, May 21, 2007 at 02:47:45PM +0200, Ola Skavhaug wrote:
> > Anders Logg skrev den 18/05-2007 følgende:
> > > Can something be done about choosing a reasonable scaling for arrows
> > > in vector plots? How about setting the maximum length to something
> > > like
> > >
> > >     D / num_vertices^(1/d)
> > >
> > > where D is the diameter of the domain, and d is the space dimension?
> > >
> > > Take a look as the DOLFIN demo in src/demo/pde/mixed-poisson and you
> > > will see what I mean... :-)
> > The test doesn't run:
> >   Using direct solver.
> >   *** Warning: UMFPACK must be installed to peform a LU solve for uBlas
> > matrices. A Krylov iterative solver will be used instead.
> >   Solving linear system of size 2112 x 2112 (uBlas Krylov solver).
> > terminate called after throwing an instance of 'std::runtime_error'
> >   what():  *** Error: Zero pivot detected in uBlas ILU preconditioner in row
> > 1600.
> > Aborted (core dumped)
> >
> > I have libufsparse (and dev package) installed, and I need these because of
> > octave2.9. libumfpack conflicts with libufsparse. What is the solution?
> >
> > Ola
>
> I don't have a good solution to the libufsparse/libumfpack problem.
> Anyone else?

There is a bug in the Debian/Ubuntu libufsparse package 
(http://bugs.debian.org/cgi-bin/pkgreport.cgi?pkg=libufsparse;dist=unstable). 
Use libumfpack instead.

The problem is that libufsparse is not linked correctly to BLAS. You 
need to supply -lblas, etc when compiling DOLFIN. Since this is a 
libufsparse problem, I didn't want to add a test for BLAS to the DOLFIN 
configure process.

Garth


> I've attached the solution (sigma.xml.gz). Let's see if it goes
> through or it's too large for the mailing list (21K).
>
> /Anders
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev