Hi all,
I am programming a problem using the techniques of step 40.
In my deal.ii serial code I run a Newton iteration process to converge
to a steady state solution by using the previous solution
values corresponding to the basis functions of each particular cell
in the assembly's cell for loop. To obtain the previous solution
values (or previous coefficients), I use the calculated solution
from the previous iteration ( equals 0 if first iteration) stored
in the public vector
Vector<double> solution
and the local to global index
std::vector<unsigned int> local_dof_indices (dofs_per_cell);
in the cell for loop
DoFHandler<3>::active_cell_iterator
cell = dof_handler.begin_active(),
endc = dof_handler.end();
for (; cell!=endc; ++cell)
{
fe_values.reinit (cell);
cell->get_dof_indices (local_dof_indices);
for (unsigned int k=0; k < dofs_per_cell; k++)
prev_coeffs[k] = solution(local_dof_indices[k]);
...
I then form Newton's method.
Would a similar call for the parallel code with the distributed
mesh as in step-40 using the processor vector
PETScWrappers::MPI::Vector locally_relevant_solution;
and the local to global index
std::vector<unsigned int> local_dof_indices (dofs_per_cell);
work in a similar manner? The indexing would be different
since we do not have the entire solution on each processor.
I was how this might work?
Thanks,
Dan
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
