Title: Dynamic frequency scaling for embedded systems with memory intensive applications

Authors: Hyoungjong Kim; Jaehyeon Jang; Moonju Park

Addresses: UbiQuoss Inc., Seongnam, Gyeonggi-Do, South Korea ' Department of Computer Science and Engineering, Incheon National University, Incheon, South Korea ' Department of Computer Science and Engineering, Incheon National University, Incheon, South Korea

Abstract: Power management of embedded systems is an important issue as they are powered from a limited energy storage. To reduce the power consumption of the processor, dynamic frequency/voltage scaling (DFS) technique has been used widely in embedded systems. In this paper, we consider the cycles per instruction (CPI) as the primary indicator of energy efficiency. On a real target hardware, we observe that memory access affects the CPI, and found that the high CPI lowers the energy efficiency and the performance gain by frequency scaling. With this observation, a new DFS algorithm based on the CPI is proposed. Experimental results show the proposed DFS algorithm provides higher energy efficiency for memory-intensive application; it reduces the energy consumption by 28.6%, without causing delay in execution. The proposed algorithm is mainly designed for memory-intensive applications, but it works well with CPU-intensive applications and mixed type of applications also.

Keywords: dynamic frequency scaling; embedded system; scheduling.

DOI: 10.1504/IJES.2018.090571

International Journal of Embedded Systems, 2018 Vol.10 No.2, pp.137 - 147

Received: 03 May 2016
Accepted: 04 Sep 2016
Published online: 22 Mar 2018 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article