Reusability for Parallel Hybrid Numerical Methods

LAKe (Linear Algebra Kernel) is a linear algebra class library developed using an object oriented approach, in order to enable very good code reuse for sequential and parallel linear algebra applications. An application implemented using LAKe can be run either in sequential or parallel mode using the same code. In this paper we propose an extension to the LAKe library, which allows making use of the same code for both sequential, and different fine grain parallelism of these methods, at the same time. This extension of LAKe allows a concurrent reuse of sequential and parallel components in the same linear algebra application. The hybrid methods are one of the well known ones which needs such properties to run. These methods are combination of several numerical techniques, or several copy of the same method parameterized differently in order to accelerate the convergence and/or to improve the accuracy of the solution of some large linear algebra problems. Some numerical experiments validate our approach.