Typically, when a user wishes to minimise the energy consumption for an application running on a handheld device, he/she may choose to set the processor speed to its slowest level. However, our study indicated that due to the processes involved in memory accesses, decreasing the CPU frequency may not always reduce the energy consumption. A critical speed has been defined as the CPU frequency, at which energy consumption can be minimised for a program. It can be used when the user wants to maximise energy saving for the device if performance is a less important issue. In this paper, an energy conservation DVFS algorithm is proposed to achieve this goal. It predicts and applies the critical speed as the target CPU frequency during the program’s execution time. The algorithm relies on a prediction equation that is constructed based on the correlation between the critical speed and the memory access rate. We have implemented the algorithm on the Android operating system. Our results show that both the energy consumption and the performance can be improved than the situation of simply selecting the lowest frequency.
목차
Abstract I. INTRODUCTION II. RELATED WORKS III. PREDICTING THE FREQUENCY FOR MINIMUM ENERGYCONSUMPTION A. The Critical Speed B. Relationship between MAR and Critical Speed C. Construction of the MAR-CSE Equation IV. IMPLEMENTATION OF THE ENERGY CONSERVATION DVFS ALGORITHM V. EVALUATION A. The Experimental Set-up B. Results of the Benchmark Programs VI. CONCLUSIONS ACKNOWLEDGMENT REFERENCES
키워드
Embedded SystemsLow Power Software DesignDVFSEnergy-conservationAndroidLinux
저자
Wen-Yew Liang [ Department of Computer Science and Information Engineering, National Taipei University of Technology, Taipei 106, Taiwan ]
Po-Ting Lai [ Department of Computer Science and Information Engineering, National Taipei University of Technology, Taipei 106, Taiwan ]
Che Wun Chiou [ Department of Computer Science and Information Engineering, Ching Yun University, Chung-Li 320, Taiwan ]