Performance and Energy Evaluation for Solving a Schrödinger-Poisson System on Multicore Processors.

In this article, a Fourier-Galerkin approach for solving a Schrödinger-Poisson system is considered. The Fourier-Galerkin approach leads to an approximation method with two steps consisting of a truncation of the Fourier-Galerkin series and the solution of the resulting ordinary differential equation with a Runge-Kutta solver. Both steps influence the numerical accuracy of the final solution as well as the performance and energy behavior. The numerical approximation software is implemented as a multi-threaded program. The exploitation of frequency scaling and the degree of parallelism provided by the number of cores may also result in different values of execution time and energy consumption. The goal of this article is to evaluate the performance behavior of the numerical approximation software based on measurements on three multicore platforms. The mutual interaction of the number of threads, the operational frequency, and the computational requirement for achieving a specific numerical accuracy of the solution are investigated. Experiments show a good scalability behavior for computing approximation solutions of different numerical accuracy.