Optimality and Improvement of Dynamic Voltage Scaling Algorithms for Multimedia Applications

Abstract

The time-varying workload for multimedia applications poses a great challenge for the efficient performance of dynamic voltage scaling (DVS) algorithms. While many DVS algorithms have been proposed for real-time applications, there does not yet exist a systematic method for evaluating the optimality of such DVS algorithms. In this paper, we propose an offline linear programming (LP) method to determine the minimum energy consumption for processing multimedia tasks under stringent delay deadlines. Based on this lower bound, we evaluate the efficiency of various existing DVS algorithms. Furthermore, we modify the LP formulation to construct an online robust sequential linear programming DVS algorithm for real-time multimedia processing. Simulation results from decoding over a wide range of video sequences shows that on average, our online algorithm consumes less than 1% more energy than the optimal lower bound while dropping only 0.1% of all scheduled decoding jobs, while the existing best algorithm consumes roughly 3% more energy at the same miss rate.

Adaptive power management system for mobile multimedia device

“…Most recently, a research [15] is investigated to deal with power consumption for multimedia application. However, this approach adopted only DVS.…”

Adaptive Coding with CPU Energy Conservation for Mobile Video Calls

“…Tan et al [5] developed a decoding workload prediction model based on the block level statistics of each MPEG frame and proposed three DVS methods. Cao et al [6] utilized robust sequential linear programming to construct a near-optimal DVS algorithm for video decoding. Sun et al [7] applied DVS according to the non-linear battery discharging state during H.264 coding.…”

A Study on Real-Time Scheduling for Low-Power Sensor Node Platforms

“…Zhong et al [4] and Kargahi et al [5] provide mathematical DVS models for computing the optimal scheduling problems. Zhang et al [6], Ishiha et al [7], and Cao et al [8] use an exact approach, based on the ILP (Integer Linear Programming) formulation. However, the work presented in [4][5][6][7][8] can be hardly applied to a sensor node platform containing a low-power 8-bit microcontroller in a practical amount of computation time.…”

“…Zhang et al [6], Ishiha et al [7], and Cao et al [8] use an exact approach, based on the ILP (Integer Linear Programming) formulation. However, the work presented in [4][5][6][7][8] can be hardly applied to a sensor node platform containing a low-power 8-bit microcontroller in a practical amount of computation time. Additionally, the work presented in [6][7][8] does not consider the workload variations, where the actual execution time of a task can be far less than its WCET.…”

“…However, the work presented in [4][5][6][7][8] can be hardly applied to a sensor node platform containing a low-power 8-bit microcontroller in a practical amount of computation time. Additionally, the work presented in [6][7][8] does not consider the workload variations, where the actual execution time of a task can be far less than its WCET. Tasks typically complete in much less time (some orders of magnitude less) than their anticipated worst-case execution times [9].…”