Hi all,
I am trying to assemble the RHS for a linear system (coming from a DG
discretization). That RHS contains a term that looks like
\int_{\Gamma_I} [[ f ]] { \nabla z } ds (f is known and z are the test
functions, so jump * average of gradient)
where Gamma_I is the union of all the interior edges in my triangulation. I've
done assemblies of local matrices for terms of this type, but this is the 1st
time I attempt this with a RHS. Here is how I go about it, using the MeshWorker
framework. I have already stored the information about f (values) in a
NamedData
object that the assembler uses (a la step-39):
template <int dim>
void MyAssembler<dim>::face(DoFInfo& dinfo1, DoFInfo& dinfo2,
CellInfo& info1, CellInfo& info2)
std::vector<double>& f_values = info1.values[0][0];
std::vector<double>& f_neighbor_values = info2.values[0][0];
Vector<double>& local_vector = dinfo1.vector(0).block(0);
Vector<double>& neighbor_vector = dinfo2.vector(0).block(0);
for (point = 0; point < n_quad_pts; ...)
for (i = 0; i < fe_v.dofs_per_cell; ...)
{
local_vector(i) += 0.5 * f_values[point] * normals[point] *
fe_v.shape_grad(i,point) * JxW[point];
neighbor_vector(i) += 0.5 * f_neighbor_values[point] * normals[point]) *
fe_v_neighbor.shape_grad(i,point) * JxW[point];
local_vector(i) -= 0.5 * f_values[point] * normals[point])
* fe_v.shape_grad(i,point) * JxW[point];
neighbor_vector(i) -= 0.5 * f_neighbor_values[point] * normals[point] *
fe_v_neighbor.shape_value(i,point) * JxW[point];
}
So basically I add the contribution containing fe_v_neighbor.shape_grad() to
the
neighbor_vector and the rest to the local vector. Is this a good approach, or
should I just sum everything in the local_vector?
Thanks!
-- Mihai
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
