Hi Lucas and Wolfgang
I have something to say on this issue because I think it might be helpful
to Lucas or other users. I know large memory usage is unavoidable in direct
solvers. But I guess Lucas' issue with MUMPS was that he did not know how
to reuse the factorization like one always does
Timo,
Thanks a lot for the reply. I would like to make my problem clear again. I
am using data_out.add_data_vector to add a solution vector. Except in my
case, the solution vector is not obtained by solving governing equations,
it is already given. The value for each dof is given in this
Jie,
Another issue related to this question is that, ApproximateMassSchur has a
vmult member function defined, so it can be used in CG solver to get the
product of its inverse with some vector, just like step-20 does.
However unlike the dealii SolverCG class, PETScWrappers::SolverCG only
Jianan,
I am using Nedelec elements to solve the Maxwell equations, which is a
complex-valued problem. When search the previous works I found that Ross Kynch
was using fe.system_to_block_index() to splitt the system into real/imaginary
parts, while QinCe was using fe.system_to_base_index().