engleski

Performance Improvement of Algebraic Multigrid Solver by Vector Sequence Extrapolation

Algebraic Multigrid Method (AMG) performance im- provement by vector sequence extrapolation is examined. Projective Forward Extrapolation (PFE), Minimal Polynomial Extrapolation (MPE) and Reduced Rank Extrapolation (RRE) are applied to the AMG resulting in a hybrid approach, vector extrapolated AMG. The impact of vector sequence extrapolation is shown to improve performance of the AMG in number of cycles and execution time, resulting in three new methods: PFE-AMG, MPE-AMG and RRE-AMG. Computational results of the application of vector extrapolated AMG to sparse matrices arising from discretization of fluid flow equations are presented showing performance improvements compared to the traditional AMG.

Algebraic Multigrid; CFD; vector extrapolated AMG

nije evidentirano