MGLab: An Interactive Environment for Multigrid Methods


                        James Bordner and Faisal Saied
                        Department of Computer Science
                  University of Illinois at Urbana-Champaign


                                  Abstract: 

MGLab is a set of Matlab functions that define an interactive environment for
experimenting with multigrid algorithms.  The package solves two-dimensional
elliptic partial differential equations using finite differences and includes
several built-in problems (Poisson, Helmholtz, discontinuous coefficient
problems and non-self adjoint problems).  A number of parameters controlling
the V-cycle can be set using a point-and-click mechanism.  The menu-based user
interface also allows a choice of several Krylov subspace accelerators (PCG,
GMRES(k), BiCG-STAB).  The package exploits Matlab's visualization and sparse
matrix features and has been structured to be extensible -- users can add
other cycles, smoothers, Krylov accelerators, problems, etc.  MGLab can also
serve as an on-line aid to teaching multigrid.