Exploration of the design space in heterogeneous architectures aware of dark silicon and using approximate computation

• Daniela Catalan UFMS
• Ricardo Ribeiro dos Santos UFMS

DOI: https://doi.org/10.5753/eradco.2020.12644

Abstract

The dark silicon problem arose with the increase in leakage current as a result of the miniaturization of the transistors. Research in order to find solutions to mitigate dark silicon has been studied, many of them proposing the heterogeneity of processing devices. However, the increase in the diversity of devices and the objectives in the definition of heterogeneous and high-performance systems, make the design of such systems more complex, requiring automated mechanisms for exploring project space aware of dark silicon. A promising solution is the use of approximate computing, in which hardware and software components use approximation instead of the precision of operations, accepting loss of output quality to improve energy efficiency and obtain performance gains. This work aims to obtain efficient solutions to the problem of exploring processor designs aware of dark silicon, using approximate computing modules as feasible elements of a heterogeneous computational system.  

Keywords: dark silicon, design space exploration, approximate computing

References

Almurib, H. A. F., Kumar, T. N., and Lombardi, F. (2016). Inexact designs for approximate low power addition by cell replacement. In Proceedings of the 2016 Conference on Design, Automation & Test in Europe, DATE ’16, page 660–665, San Jose, CA, USA. EDA Consortium.

Chippa, V., Mohapatra, D., Roy, K., Chakradhar, S., and Raghunathan, A. (2014). Scalable effort hardware design. Very Large Scale Integration (VLSI) Systems, IEEE Transactions on, 22:2004–2016.

Gorantla, A. and Deepa, P. (2019). Design of approximate subtractors and dividers for error tolerant image processing applications. Journal of Electronic Testing, pages 1–7.

Gupta, V., Mohapatra, D., Raghunathan, A., and Roy, K. (2013). Low-power digital signal processing using approximate adders. IEEE Trans. on CAD of Integrated Circuits and Systems, 32(1):124–137.

Jiang, H., Han, J., and Lombardi, F. (2015). A comparative review and evaluation of approximate adders. In Proceedings of the 25th Edition on Great Lakes Symposium on VLSI, GLSVLSI ’15, page 343–348, New York, NY, USA. Association for Computing Machinery.

Muthulakshmi, S., Dash, C., and Prabaharan, S. (2018). Memristor augmented approximate adders and subtractors for image processing applications: An approach. AEU - International Journal of Electronics and Communications, 91.

Santos, R., Sonohata, R., Krebs, C., Catelan, D., Duenha, L., Segovia, D., and Santos, M. (2019). Explorac¸ao do projeto de sistemas baseados em gpu ciente de dark silicon. In Anais Principais do XX Simposio em Sistemas Computacionais de Alto Desempenho, pages 358–369, Porto Alegre, RS, Brasil. SBC.

Sassi, A. B. (2013). Projeto de uma ULA de inteiros e de baixo consumo em tecnnologia CMOS. PhD thesis, Escola de Engenharia de Sao Carlos da Universidade de São Paulo.

Yang, Z., Jain, A., Liang, J., Han, J., and Lombardi, F. (2013). Approximate xor/xnorbased adders for inexact computing. Proceedings of the IEEE Conference on Nanotechnology, pages 690–693.