Custo energético em computação de alto desempenho

• Paulo N. M. dos Anjos UNIFESP
• Alvaro L. Fazenda UNIFESP

DOI: https://doi.org/10.5753/eradsp.2023.232682

Abstract

The development of new algorithms usually focuses on improving performance, with little consideration given to environmental impact and energy cost associated with their execution. However, this topic has been receiving increased attention recently. This work aims to objectively demonstrate the energy consumption of commonly found programming patterns in computationally intensive scientific programs. The energy consumption will be measured using software through the RAPL interface, while other performance measures will utilize the number of operations executed and elapsed time. Variations will be made in the number of threads used and compilation options to identify the impacts these changes have on energy efficiency.

References

Anwar, M., Qadri, S., and Sattar, A. (2013). Green computing and energy consumption issues in the modern age. IOSR Journal of Computer Engineering, 12(6):91–98.

Malony, A. D. (2014). Stencil pattern. Disponível em: [link]. Acesso em: 04/01/2022.

Pandruvada, S. (2014). Running average power limit. Disponível em: [link]. Acesso em: 06/01/2022.

Rotem, E., Naveh, A., Ananthakrishnan, A., Weissmann, E., and Rajwan, D. (2012). Power-management architecture of the intel microarchitecture code-named sandy bridge. IEEE Micro, 32(2):20–27.

Salama, M. (2020). Green computing, a contribution to save the environment. Disponível em: [link]. Acesso em: 06/01/2022.

University of California (2021). Parallel computation patterns (reduction). Disponível em: [link]. Acesso em: 04/01/2022.