Algebraic-geometric multigrid methods of domain decomposition
V.P. Il’in1,2
1Institute of Computational Mathematics and Mathematical Geophysics of Siberian Branch of Russian Academy of Sciences, Novosibirsk, Russia
2Novosibirsk State Technical University, Novosibirsk, Russia
Keywords: preconditioned Krylov methods, multidimensional problems, domain decomposition, multigrid approaches, incomplete factorization, diagonal compensation, parallelization of algorithms
Abstract
Some iterative processes in Krylov subspaces are considered for solving systems of linear algebraic equations (SLAE) with high-order sparse matrices that arise in grid approximations of multidimensional boundary value problems. The SLAE are preconditioned by a uniform combined method that includes domain decomposition and recursive application of a two-grid algorithm, which are implemented by forming block-tridiagonal algebraic and grid structures inverted by using incomplete factorization and diagonal compensation. For some Stieltjes systems, stability and convergence of iterations are studied. Parallelization and generalization of the methods to wider classes of relevant practical problems are discussed.
|
