Efficient Setup Algorithms for Parallel Algebraic Multigrid

The experimental results motivate the design of new coarsening algorithms to improve the performance of coarse-grid selection itself. A new algorithm labeled Bucket Sorted Independent Sets (BSIS) is developed and contributes two major advances. First, the cost of selecting independent sets while coarsening is substantially less expensive, with experiments demonstrating 23% savings over CLJP-c. Second, theory is developed proving that all generalized forms of the coarsening algorithms studied in this thesis using the same selection and update parameters choose identical coarse grids, given the same initial weights. The theory is powerful because it provides insight and enables the development of more efficient algorithms without affecting convergence properties.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Alber, David Michael

Contributors dc:contributor

• Olson, Luke N.

• Computer Science

Language dc:language

eng

Identifier

(MiAaPQ)AAI3269829

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/81758