Parallel Architectures and the Simulation of N-Body Systems: An Analysis

• Lucas Menduiña Ramos Evangelista UFRRJ
• Gabriel Marinho de Souza UFRRJ
• Juliana Nascente UFRRJ
• Marcelo Zamith UFRRJ

DOI: https://doi.org/10.5753/eradrj.2024.4338

Abstract

The N-body problem, widely studied in astrophysics, involves predicting the motion of celestial bodies under the influence of their mutual gravitational interactions. This study proposes an analisys of two variants of the problem, using high-performance computing, leveraging parallelism on both CPUs and GPUs. The implementations utilize efficient load balancing techniques, based on data and task parallelization, with a focus on OpenMP and Nvidia CUDA. Parallelizing the gravitational force calculations and mass distribution results in significant speedup and reduced execution time, empirically validated and presented graphically. In addition to offering an efficient alternative to complex simulations, this work has potential applications in other domains involving large-scale data and intensive computational tasks.

References

Clua, E. W. G. and Zamith, M. P. (2015). Programming in cuda for kepler and maxwell architecture. Revista de Informática Teórica e Aplicada, 22(2):233–257.

Gangavarapu, T., Pal, H., Prakash, P., Hegde, S., and Geetha, V. (2019). Parallel openmp and cuda implementations of the n-body problem. In: Misra, S., et al. Computational Science and Its Applications – ICCSA 2019. ICCSA 2019. Lecture Notes in Computer Science(), vol 11619. Springer.

Moore, G. (1965). Moore’s law. Electronics Magazine, 38(8):114.

Patterson, D. A. and Hennessy, J. L. (2013). Computer Organization and Design, Fifth Edition: The Hardware/Software Interface. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 5th edition.