¹University of Shanghai for Science and Technology, School of Optical-Electrical and Computer Engineering
²2University of ShangHai for Science and Technology, , School of Management, Shanghai, China

---

Abstract—Energy efficiency is a major concern in the General Programming on Graphic Process Unit. Recent research focus on the measurement approach and energy optimization of Graphic Process Unit. Few studies provide insight to where and how power is consumed from the program perspective. The aim of this research was to build power consumption model to estimate the energy consumption for the application programmers. Program slicing was used to decompose the programs into slice set. The program slice as basic unit was to measure and analyze the program power consumption. We consider the computation intensity and the number of active SMs that have directly impact on energy consumption. Aiming to the sparseness-branch and denseness-branch programs, two power consumption prediction models were proposed. The experimental results show that the average relative error of the two prediction models are less than 6 percent. We conclude that the power consumption prediction models can effectively estimate the energy consumption of applications.

Index Terms—Power Consumption Model, GPU Computing, CUDA, Program Slicing

[PDF]

Cite: Haifeng Wang and Qingkui Chen, "Power Estimating Model and Analysis of General Programming on GPU," Journal of Software vol. 7, no. 5, pp. 1164-1170, 2012.