Open Multi Processing (OpenMP) of Gauss-Jordan Method for Solving System of Linear Equations

Abstract

Solving systems of linear equations is probably one of the most scientific applications of linear algebra and direct-based Gauss-Jordan method as a classical kernel of large system of linear equations has become the focus of research. This paper presents an OpenMP pipeline implementation of Gauss-Jordan method and the corresponding performance model. Then, we conduct an experimental evaluation of the pipeline implementation in comparison with the two other naive parallel versions of Gauss-Jordan method such as row block and row cyclic distribution on a multicore platform. From the experiments are obtained show that the proposed pipelined implementation is a good solution for solving large system of linear equations. Finally, the experimental results of the pipeline implementation confirm the proposed performance model.