A Method for Solving Large Sparse Unsymmetric Linear Equations

摘要

The computational cost of large sparse unsymmetric linear equations has brought a challenge to engineering and scientific fields in recent years. In this paper, we present a parallel algorithm for solving large sparse unsymmetric linear equations. We multiply the unsymmetric linear coefficient matrix by its transpose to transform the large sparse unsymmetric matrix into a symmetric one, in this way one can use existing scalable hierarchical parallel algorithms designed for symmetric problems to solve it. In addition, we parallel this process, and the decomposition method designed in this paper insure the parallel performance of the algorithm. Numerical results demonstrate that this algorithm possesses excellent parallel efficiency and good expansibility.