Performance and Energy Evaluation of Parallel Particle Simulation Algorithms for Different Input Particle Data

Abstract

Particle simulations are popular methods for the simulation of applications from a wide range of sciences, including astrophysics, biology or chemistry. Usually, these applications require a large number of simulation steps, each of which computes a change of the entire particle system. Depending on the number of simulation steps and also the size and structure of the specific particle system, the computation time can be quite large and the exploitation of parallel architectures is usually necessary. In this article, we investigate the performance and energy consumption for different particle simulation methods and distinguish different input particle data. The investigations are done for the particle simulation methods from the ScaFaCoS library and use the various input data of homogeneous or inhomogeneous nature. Experiments are performed on multicore systems.

A performance- and energy-oriented extended tuning process for time-step-based scientific applications

“…This approach does not consider the energy consumption and does not optimize the number of threads. Runtime and energy consumption of particle simulations are examined in [16,22]. The tuning potential with respect to performance for particle simulation methods and solving differential equations is examined in [20].…”