Evaluation of Approximate Computing Techniques for Power Reduction on FPGAs

Echavarria Gutiérrez JA, Schütz K, Becher A, Wildermann S, Teich J (2018)


Publication Language: English

Publication Type: Conference contribution, Original article

Publication year: 2018

Event location: Swissôtel Bremen DE

Open Access Link: https://www12.cs.fau.de/downloads/echavarria/pub/Evaluation_of_Approximate_Computing_Techniques_for_Power_Reduction_on_FPGAs.pdf

Abstract

Approximate computing allows to tackle conflicting objectives, such as power and accuracy of computations. In this paper we first describe how knowledge of stimuli's specific features can help in quantifying and improving power savings by means of approximate computing. We investigate FPGA implementations of several approximate circuits and compare their power consumption with non-approximating versions. In particular, we study approximate arithmetics and a clock-gate based technique called memoization. Moreover, we compare the accuracy of estimation techniques for power consumption evaluation versus real measurements under controlled environments. We also experimentally quantify the relationship between switching activity and power consumption. Two important results are concluded from our investigations: (1) Approximate arithmetics do not necessarily consume less power than conventional circuits, whereas memoization techniques can in fact reduce power consumption. (2) Simulation-based power evaluation for approximate FPGA implementations can reach fidelity values up to about 90% in input-dependent power characteristics. Yet, to evaluate absolute savings, measurements are needed.

Authors with CRIS profile

Related research project(s)

How to cite

APA:

Echavarria Gutiérrez, J.A., Schütz, K., Becher, A., Wildermann, S., & Teich, J. (2018). Evaluation of Approximate Computing Techniques for Power Reduction on FPGAs. In Proceedings of the AxC18: 3rd Workshop on Approximate Computing. Swissôtel Bremen, DE.

MLA:

Echavarria Gutiérrez, Jorge Alfonso, et al. "Evaluation of Approximate Computing Techniques for Power Reduction on FPGAs." Proceedings of the AxC18: 3rd Workshop on Approximate Computing, Swissôtel Bremen 2018.

BibTeX: Download