Parallelizing the Computation of Green Functions for Computational Electromagnetism Problems

Abstract

Green functions are used in various fields to solve non homogeneous integral equations with boundary conditions. In some cases it is necessary to obtain these functions in real time or they are used for big problems with a large execution time which should be reduced. To do so, efficient algorithms for the available computational systems must be developed. With the evolution of technology, the computational systems are now parallel, with multicore laptops or desktops with programmable graphic processing units, and with clusters or supercomputers composed by multicore nodes. In this paper, algorithms for Green functions using different parallelism paradigms are developed and compared. The Green functions used are from computational electromagnetism, and important reductions in the execution time are obtained for typical problems with the implemented algorithms.