Técnicas de precondicionamiento en paralelo.

Abstract

The investigation's main objective was to design and implement a parallel preconditioner technique which could help solve linear equation systems required for the numerical solution of partial di erential equations, systems whose common characteristics are their big size and the fact that the related coe cient matrix is sparse. The parallel preconditioner technique designed and implemented is based in the incomplete LU factorization (ILU), and the preconditioner was applied to problems found in The University of Florida Sparse Matrix Collection. The preconditioner was used in unsymetric matrices and in non-positive de nite symetric matrices, and the best results were found in cases of non-positive de nite symetric matrices.