Energy estimation for 32-bit microprocessors

Estimation of software power consumption is becoming one of the major problems for many embedded applications. The paper presents a novel approach to compute the energy of an instruction set, through a suitable functional decomposition of the activities involved during instruction execution. One of the main advantages of this approach is the capability to predict the power figures of the overall instruction-set starting from a small subset. A formal discussion on the statistical properties of the model is included, together with its application on five commercial 32-bit microprocessors.