← Back to team overview

dolfin team mailing list archive

  1. dolfin team
  2. Mailing list archive
  3. Message #02919

Re: normalized vectors; got SIGSEGV

  Thread Previous • Date Previous • • Thread Next 


Anders Logg wrote:

On Mon, Jul 24, 2006 at 06:56:32PM +0200, Dag Lindbo wrote:

Hi all,

The Function class is causing some headaches for me, as I try to
normalize 2-vectors on my mesh. I have a form that defines the
computation of a gradient, so in my solver i have a vactor-values
function, normal_phi (whose components are the gradient vectors at each
vertex). So far it's OK. Now I loop over all vertices and manually
normalize each vector in normal_phi. This works too.

But then the function, normal_phi, goes into a another form, as a
Function("vector-element"). Upon this call I get a sig11: Segmentation
Violation. If I do no normalization I get no violation! So I wonder what
has happened. The code is very short, so I hope someone could look at
it:


unsigned int Nv = FEM::size(mesh, element);
(...)
grad_pde.solve(norml_phi); //When I write norml_phi[0] and norml_phi[1]
to file they look great
Function grad_phi = norml_phi[0]; Vector& norml = grad_phi.vector();
Why are you taking a sub-function here? For the below code to make sense, should you use 

Vector& norml = norml_phi.vector();
That said, the above might work since as far as I recall, the sub-function grad_phi has a copy of the entire vector from norml_phi and 
grad_phi.vector() returns a reference to the entire vector (I think).



for(VertexIterator vertex(mesh); !vertex.end(); ++vertex)
 {
  unsigned int vid = vertex->id();
  real gpx = norml( vid );
  real gpy = norml( vid + Nv);
  real p = std::sqrt( gpx*gpx+gpy*gpy );

  norml( vid ) = gpx/p;
  norml( vid + Nv) = gpy/p;
 }

norml_phi.attach(norml); // I write to VTK-file here and it looks
perfect
(...)

kappa_pde.solve(kappa); // < SIG 11
Not sure why you get the error message, but assuming that the gradients are not the primal unknown in your problem I would recommend computing the term 

   normal_grad = gpx*gpx + gpy*gpy
directly using FFC, then take the square root. This will generalise more easily to any type of element. 

Garth







Best regards,
Dag Lindbo

_______________________________________________
DOLFIN-dev mailing list
DOLFIN-dev@xxxxxxxxxx
http://www.fenics.org/mailman/listinfo/dolfin-dev


Just back from my vacation...

Did you solve this or is this still a problem?

/Anders
_______________________________________________
DOLFIN-dev mailing list
DOLFIN-dev@xxxxxxxxxx
http://www.fenics.org/mailman/listinfo/dolfin-dev


--

Dr. Garth N. Wells
Faculty of Civil Engineering and Geosciences
Delft University of Technology
Stevinweg 1
2628 CN Delft
Netherlands

tel.   +31 15 278 7922
fax.   +31 15 278 6383
e-mail g.n.wells@xxxxxxxxxx
url    http://www.mechanics.citg.tudelft.nl/~garth

References

  Thread Previous • Date Previous • • Thread Next