Eigenanalysis of finite element 3D flow models by parallel Jacobi-Davidson

Computing some eigenpairs of a Finite Element (FE) flow model is an important task. Parallel computations are unavoidable, in order to accurately analyze in an acceptable time large structures, e.g. regional groundwater systems, by fully 3D FE models. The eigenpairs can be efficiently computed by Jacobi–Davidson (JD) algorithm, provided it is efficiently preconditioned. Parallelization of JD exploiting FSAI preconditioning technique wascompleted under a distributed memory paradigm, by a Fortan 90 code performing MPI calls. In this paper we show that our code allows for the fast eigensolution of huge real–life flow problems on a parallel IBM SP4 machine.