Theoretical bounds for algebraic multigrid performance: review and analysis

Authors


SUMMARY

Algebraic multigrid methods continue to grow in robustness as effective solvers for the large and sparse linear systems of equations that arise in many applications. Unlike geometric multigrid approaches, however, the theoretical analysis of algebraic multigrid is less predictive of true performance. Multigrid convergence factors naturally depend on the properties of the relaxation, interpolation, and coarse-grid correction routines used, yet without the tools of Fourier analysis, optimal and practical bounds for algebraic multigrid are not easily quantified. In this paper, we survey bounds from existing literature, with particular focus on the predictive capabilities of the theory, and provide new results relating existing bounds. We highlight the impact of these theoretical observations through several model problems and discuss the role of theoretical bounds on practical performance. Copyright © 2014 John Wiley & Sons, Ltd.

Ancillary