Electronic Transactions on Numerical Analysis

Previously we examined the black box multigrid approach to systems of equations. The approach was a direct extension of the methodology used for scalar equations; that is, interpolation and residual weighting were operator induced, and coarsening employed a Galerkin strategy. The application was to standard coarsening of the unknowns. In this paper we consider a semicoarsening approach and find that there are a few differences in what is generally effective. 1. Introduction. In [3] we extended to systems of equations the ideas contained in [1] and [2]. More specifically, let us consider multigrid with standard coarsening on a rectangular grid of points; that is, the coarse grid offspring of the grid G