Parallelization of Monte Carlo Based Poisson Solver with Coarray Fortran
Nevşan Şengil, Özgür Tümüklü · International Conference on Applied and Computational Mathematics · 2012
Poisson’s equations prevail in many scientific problems. Although very different techniques exist to solve Poisson's equation, because of its tendency to easy parallelization, we focused on the Monte Carlo method. The aim of parallelization is to increase the efficiency and reduce the computation time. Among the many parallelization paradigms, such as OpenMP, MPI and PGAS, we adopted PGAS based coarray Fortran (CAF). Poisson's equation is solved for the non-rectangular domain on both multiprocessors and multicore platforms. The efficiency figures show that CAF is a very capable parallelization tool in addition to the portability and maintainability qualities.