Energy-demand estimation of embedded devices using deep artificial neural networks

Hönig T, Herzog B, Schröder-Preikschat W (2019)

Publication Type: Conference contribution

Publication year: 2019

Publisher: Association for Computing Machinery

Book Volume: Part F147772

Pages Range: 617-624

Conference Proceedings Title: Proceedings of the ACM Symposium on Applied Computing

Event location: Limassol CY

DOI: 10.1145/3297280.3297338

Abstract

The need for high performance in embedded devices grows at a breathtaking pace. Embedded processors that satisfy the hunger for superlative processing power share a common issue: the increasing performance leads to growing energy demands during operation. As energy remains a limited resource to embedded devices, it is critical to optimise software components for low power. Low-power software needs energy models which, however, are increasingly difficult to create as to the complexity of today's devices. In this paper we present a black-box approach to construct precise energy models for complex hardware devices. We apply machine-learning techniques in combination with fully automatic energy measurements and evaluate our approach with an ARM Cortex platform. We show that our system estimates the energy demand of program code with a mean percentage error of 1.8 % compared to the results of energy measurements.

Authors with CRIS profile

Timo Hönig Lehrstuhl für Informatik 4 (Verteilte Systeme und Betriebssysteme) Benedict Herzog Lehrstuhl für Informatik 4 (Verteilte Systeme und Betriebssysteme) Wolfgang Schröder-Preikschat Lehrstuhl für Informatik 4 (Verteilte Systeme und Betriebssysteme)

How to cite

APA:

Hönig, T., Herzog, B., & Schröder-Preikschat, W. (2019). Energy-demand estimation of embedded devices using deep artificial neural networks. In Proceedings of the ACM Symposium on Applied Computing (pp. 617-624). Limassol, CY: Association for Computing Machinery.

MLA:

Hönig, Timo, Benedict Herzog, and Wolfgang Schröder-Preikschat. "Energy-demand estimation of embedded devices using deep artificial neural networks." Proceedings of the 34th Annual ACM Symposium on Applied Computing, SAC 2019, Limassol Association for Computing Machinery, 2019. 617-624.

BibTeX: Download