Statistical Derivation of an Accurate Energy Consumption Model for Embedded Processors

The energy consumption of software is becoming an increasingly important issue in designing mobile embedded systems where batteries are used as the main power source. As a consequence, recently, a number of promising techniques have been proposed to optimize software for reduced energy consumption. Such low-power software techniques require an energy consumption model that can be used to identify the factors contributing to the overall energy consumption. We propose a technique to derive an accurate energy consumption model by abstracting the energy behavior of the target processor. The proposed approach combines empirical measurement with a statistical analysis technique to approximate the actual energy consumption, whose result is a model equation that can be used to estimate software energy consumption. The model equation also provides insightful information that can be used in program optimization for low energy, by identifying the factors affecting the energy consumption of software. Experimental results show that the model equation can accurately estimate the energy consumption of a random instruction sequence, with an average error of 2.5 %. An earlier version of this paper appeared in the Proceedings of the ACM SIGPLAN Workshop on Languages, Compilers, and Tools for Embedded Systems, 2001 [16]. School of Computer Science and Engineering, Seoul National University, Seoul 151-742, Korea. (TEL) +82-2-880-1831, (FAX) +82-2-885-7296, (EMAIL) [email protected], [email protected], [email protected]. Dept. of Information Technology, Uppsala University, SE-751 05 Uppsala, Sweden. (TEL) +46-18-4713172, (FAX) +46-18550225, (EMAIL) [email protected].