Dhrystone million instructions per second–based task offloading from smartwatch to smartphone

Abstract

Wearable devices such as smartwatches are critical to energy consumption due to their small battery size. To reduce possible energy consumption for high computational tasks, task offloading is a solution, but smartwatches are usually connected to smartphones that are also exposed to energy-saving issues. To resolve this problem, we propose a task offloading model that compares the energy efficiency of two cases: offloading or no offloading. When a task is offloaded, energy of a smartphone is sacrificed instead of a smartwatch, and additional energy is used for the communication. To derive the condition of an offloading, the size of a task could be considered, but it is impossible to estimate accurate energy due to various computational specifications of smartphones as well as smartwatches. To generalize the task size for different CPU conditions, we propose to use Dhrystone million instructions per second. The proposed model was evaluated through our testbed and confirmed that it shows a reasonable offloading condition for both the smartphone and the smartwatch.

Design and Implementation of a Usability-Framework for Smartwatches

“…Research design adapted from Peffers et al [34] Page 6634 environments [31]. Due to the high level of miniaturization, the limited computing and battery capacity get into the focus [23]. In addition, the connectivity of a smartwatch to wide area networks cannot be assured for any point in time.…”

Energy-Efficient Control of Mobile Processors Based on Long Short-Term Memory

“…The rν l_max , rν l_min values are then obtained by running Dhrystone benchmark. In addition, if it is assumed that the performance of each core changes linearly according to the frequency [33], then erν l_t can be defined as in Eq. (8), using linear interpolation when the frequency of the l-th processor core at time t is ν l_t .…”