Data type dependent energy consumption estimation

Abstract

Limited energy sources in embedded systems have driven the developers to search for new ways to optimise and estimate the energy consumption. Although energy estimations are usually conducted on instruction level models we have proposed a method based on the energy consumption of a C-operation. As some energy estimation techniques are meant to be applied in the very early of the design process, our proposed method is most applicable when a substantial part of the functional program code has already been programmed. This paper will show that different data types have significant impact on the overall energy consumption on microcontrollers. We apply our energy estimation methodology on different data types using a simple image processing algorithm and show that the estimate error is less than 8% with one exception.