Premium
Parallel block–selective algebraic multigrid in foam‐extend
Author(s) -
Uroić Tessa,
Jasak Hrvoje
Publication year - 2019
Publication title -
pamm
Language(s) - English
Resource type - Journals
ISSN - 1617-7061
DOI - 10.1002/pamm.201900424
Subject(s) - multigrid method , coefficient matrix , galerkin method , block (permutation group theory) , algebraic equation , matrix (chemical analysis) , system of linear equations , mathematics , algebraic number , projection (relational algebra) , finite volume method , finite element method , algorithm , mathematical optimization , mathematical analysis , materials science , physics , geometry , mechanics , partial differential equation , thermodynamics , eigenvalues and eigenvectors , quantum mechanics , nonlinear system , composite material
We present an extension of a well–known classic algebraic multigrid algorithm, where coarsening is based on selection of equations, for an implicitly coupled pressure–velocity system discretised by the finite volume (FV) method. The resulting coefficient matrix of the linear system is point–ordered, and there are 4 equations for each FV cell: three components of velocity and pressure. A significant challenge is posed in terms of parallelisation of the sequential coarsening process and calculation of coarse level matrix, using the Galerkin projection. In this paper we present the parallelisation strategy and performance of the block–selection AMG.