Hi, I am solving several problems simultaneously in parallel, similar to the question:
http://fenicsproject.org/qa/6089/solve-problems-parallelly-with-fenics-which-relative-with
Is it possible to reuse a factorization in parallel? Consider the code:
solver = LUSolver()
solver.set_operator(A)
solver.parameters["reuse_factorization"] = True
This obviously works when using one processor, however when using multiple processors the factorization is computed at each rank. So if I do not have as many processors as problems that I wish to solve then I will basically have to wait for the additional factorizations to be computed.
Alternatively: Does anyone know what pack is used for computing the factorization, such that I might use that for equally efficient LU factorization and then transfer the L and U matrices.
