4/29/2009
Written by
Researchers at the University of Illinois department of computer science have released a Python/C++ package of Algebraic Multigrid solvers. The new framework, developed by computer science professor Luke Olson and his students Nathan Bell and Jacob Schroeder, provides a more extensible and accessible toolkit to design new AMG methods.
Algebraic Multigrid is a multilevel technique for solving large-scale linear systems with optimal or near-optimal efficiency. Unlike geometric-based multigrid approaches, AMG requires little or no geometric information about the underlying problem and develops a sequence of coarser grids directly from the input matrix. This feature is especially important for problems discretized on unstructured meshes and irregular grids.
Scalable solvers are an important ingredient in the computational simulation of a wide range of physical problems. PyAMG is currently being used in finite volume and finite element software.
The team was inspired to create a more usable, flexible, and extensible approach for developing AMG methods to allow for quick testing and prototyping of additional functionality and algorithms.
The tool is available for download and use at www.pyamg.org.