Raymond S. Tuminaro

ML 3.0 Smoothed Aggregation User's Guide

ML is a multigrid preconditioning package intended to solve linear systems of equations Ax = b where A is a user supplied n x n sparse matrix, b is a user supplied vector of length n and x is a vector of length n to be computed. ML should be used on large sparse linear systems arising from partial differential equation (PDE) discretizations. While technically any linear system can be considered, ML should be used on linear systems that correspond to things that work well with multigrid methods (e.g. elliptic PDEs). ML can be used as a stand-alone package or to generate preconditioners for a traditional iterative solver package (e.g. Krylov methods). We have supplied support for working with the Aztec 2.1 and AztecOO iterative package [16]. However, other solvers can be used by supplying a few functions. This document describes one specific algebraic multigrid approach: smoothed aggregation. This approach is used within several specialized multigrid methods: one for the eddy current formulation for Maxwells equations, and a multilevel and domain decomposition method for symmetric and nonsymmetric systems of equations (like elliptic equations, or compressible and incompressible fluid dynamics problems). Other methods exist within ML but are not described in this document. Examples are given illustrating the problem definition and exercising multigrid options.

An Improved Algebraic Multigrid Method for Solving Maxwell's Equations

We propose two improvements to the Reitzinger and Schoberl algebraic multigrid (AMG) method for solving the eddy current approximations to Maxwells equations. The main focus in the Reitzinger/Schoberl method is to maintain null space properties of the weak

A New Petrov-Galerkin Smoothed Aggregation Preconditioner for Nonsymmetric Linear Systems

\nabla \times \nabla \times

Towards a scalable fully-implicit fully-coupled resistive MHD formulation with stabilized FE methods

operator on coarse grids. While these null space properties are critical, they are not enough to guarantee h-independent convergence of the overall multigrid method. We illustrate how the Reitzinger/Schoberl AMG method loses h-independence due to the somewhat limited approximation property of the grid transfer operators. We present two improvements to these operators that not only maintain the important null space properties on coarse grids but also yield significantly improved multigrid convergence rates. The first improvement is based on smoothing the Reitzinger/Schoberl grid transfer operators. The second improvement is obtained by using higher order nodal interpolation to derive the corresponding AMG interpolation operators. While not completely h-independent, the resulting AMG/CG method demonstrates improved convergence behavior while maintaining low operator complexity.

Performance of a parallel algebraic multilevel preconditioner for stabilized finite element semiconductor device modeling

We propose a new variant of smoothed aggregation (SA) suitable for nonsymmetric linear systems. The new algorithm is based on two key generalizations of SA: restriction smoothing and local damping. Restriction smoothing refers to the smoothing of a tentative restriction operator via a damped Jacobi-like iteration. Restriction smoothing is analogous to prolongator smoothing in standard SA and in fact has the same form as the transpose of prolongator smoothing when the matrix is symmetric. Local damping refers to damping parameters used in the Jacobi-like iteration. In standard SA, a single damping parameter is computed via an eigenvalue computation. Here, local damping parameters are computed by considering the minimization of an energy-like quantity for each individual grid transfer basis function. Numerical results are given showing how this method performs on highly nonsymmetric systems.

A new smoothed aggregation multigrid method for anisotropic problems

This paper explores the development of a scalable, nonlinear, fully-implicit stabilized unstructured finite element (FE) capability for 2D incompressible (reduced) resistive MHD. The discussion considers the implementation of a stabilized FE formulation in context of a fully-implicit time integration and direct-to-steady-state solution capability. The nonlinear solver strategy employs Newton-Krylov methods, which are preconditioned using fully-coupled algebraic multilevel preconditioners. These preconditioners are shown to enable a robust, scalable and efficient solution approach for the large-scale sparse linear systems generated by the Newton linearization. Verification results demonstrate the expected order-of-accuracy for the stabilized FE discretization. The approach is tested on a variety of prototype problems, including both low-Lundquist number (e.g., an MHD Faraday conduction pump and a hydromagnetic Rayleigh-Bernard linear stability calculation) and moderately-high Lundquist number (magnetic island coalescence problem) examples. Initial results that explore the scaling of the solution methods are presented on up to 4096 processors for problems with up to 64M unknowns on a CrayXT3/4. Additionally, a large-scale proof-of-capability calculation for 1 billion unknowns for the MHD Faraday pump problem on 24,000 cores is presented.

Toward an h-Independent Algebraic Multigrid Method for Maxwell's Equations

In this study results are presented for the large-scale parallel performance of an algebraic multilevel preconditioner for solution of the drift-diffusion model for semiconductor devices. The preconditioner is the key numerical procedure determining the robustness, efficiency and scalability of the fully-coupled Newton-Krylov based, nonlinear solution method that is employed for this system of equations. The coupled system is comprised of a source term dominated Poisson equation for the electric potential, and two convection-diffusion-reaction type equations for the electron and hole concentration. The governing PDEs are discretized in space by a stabilized finite element method. Solution of the discrete system is obtained through a fully-implicit time integrator, a fully-coupled Newton-based nonlinear solver, and a restarted GMRES Krylov linear system solver. The algebraic multilevel preconditioner is based on an aggressive coarsening graph partitioning of the nonzero block structure of the Jacobian matrix. Representative performance results are presented for various choices of multigrid V-cycles and W-cycles and parameter variations for smoothers based on incomplete factorizations. Parallel scalability results are presented for solution of up to 10^8 unknowns on 4096 processors of a Cray XT3/4 and an IBM POWER eServer system.

A General Interpolation Strategy for Algebraic Multigrid Using Energy Minimization

A new prolongator is proposed for smoothed aggregation (SA) multigrid. The proposed prolongator addresses a limitation of standard SA when it is applied to anisotropic problems. For anisotropic problems, it is fairly standard to generate small aggregates (used to mimic semi-coarsening) in order to coarsen only in directions of strong coupling. Although beneficial to convergence, this can lead to a prohibitively large number of non-zeros in the standard SA prolongator and the corresponding coarse discretization operator. To avoid this, the new prolongator modifies the standard prolongator by shifting support (non-zeros within a prolongator column) from one aggregate to another to satisfy a specified non-zero pattern. This leads to a sparser operator that can be used effectively within a multigrid V-cycle. The key to this algorithm is that it preserves certain null space interpolation properties that are central to SA for both scalar and systems of partial differential equations (PDEs). We present two-dimensional and three-dimensional numerical experiments to demonstrate that the new method is competitive with standard SA for scalar problems, and significantly better for problems arising from PDE systems.

A New Approximate Block Factorization Preconditioner for Two-Dimensional Incompressible (Reduced) Resistive MHD

We propose a new algebraic multigrid (AMG) method for solving the eddy current approximations to Maxwells equations. This AMG method has its roots in an algorithm proposed by Reitzinger and Schoberl. The main focus in the Reitzinger and Schoberl method is to maintain null-space properties of the weak

A new perspective on strength measures in algebraic multigrid

\nabla \times \nabla \>{\times}