Hello,
I am solving a linear system for a fluid dynamics problem of which I need to extract even some eigensolutions. I used solve(A, x, b, "lu") in c++ to solve it and I got the right solution. When I invoked SLEPcEigensolver with shift-and-invert option I received PETSc error messages. It complained about zero pivots in LU factorization.
The matrix could be singular, but in this case I can't understand how Fenics solved the system, maybe a different kind of pivoting. I could not find the objects where LU factorization and system solution is actually implemented in Fenics, to have a better insight in what is happening.
Thank you a lot.