Performance, power consumption, and dark silicon estimations in GPGPU projects

• Gislayne G. Damasceno UFMS
• Laura B. Ludgero UFMS
• Samuel S. Rodrigues UFMS
• Ricardo R. dos Santos UFMS
• Liana D. Duenha UFMS

DOI: https://doi.org/10.5753/eradco.2024.4530

Abstract

This paper presents an extension of the MultiExplorer tool for the analysis and exploration of the GPGPU designs. New simulators were integrated to characterize the performance, area, and power consumption of six GPU models running CUDA and Rodinia applications. Additionally, a core database was modeled to explore different scenarios.

References

Devigo, R., Duenha, L., Azevedo, R., and Santos, R. (2015). Multiexplorer: A tool set for multicore system-on-chip design exploration. In 2015 IEEE 26th International Conference on Application-specific Systems, Architectures and Processors (ASAP), pages 160–161.

MOORE, G. (1998). Cramming more components onto integrated circuits. Proceedings of the IEEE, 86(1):82–85.

SANTOS, R., DUENHA, L., SILVA, A., SOUSA, M., TEDESCO, L., MELGAREJO, J., SANTOS, T., AZEVEDO, R., and MORENO, E. (2018). Dark-silicon aware design space exploration. Journal of Parallel and Distributed Computing, 120:295–306.

Sonohata, R., Arigoni, D. C. A., Fernandes, E. R., Ribeiro dos Santos, R., and Dessandre Duenha, L. (2023). Performance predictors for graphics processing units applied to dark-silicon-aware design space exploration. Concurrency and Computation: Practice and Experience, 35(17):e6877.