An Automated Power Emulation Framework for Embedded Software - Detecting Power-Critical Code Regions and Optimizing Software-Induced Power Consumption Peaks

摘要

Software-induced power consumption peaks can lead to internal supply voltage drops that threaten the reliability of power-constrained mobile systems such as RF-powered smart cards. In this paper we investigate the use of the power emulation technique for automatically detecting power consumption peaks and hence supply voltage drops caused by embedded software applications. Within an automated framework this information is employed to identify the power-critical source code regions causing these power peaks. Afterwards, power optimization strategies that are available on the given system, such as frequency scaling or non-functional instruction (NFI) insertion, are automatically applied to the source code in order to mitigate the impact of the power peaks. We illustrate the applicability of our automated approach on an RF-powered smart card system and present the benefits of these fine-grained code optimizations. With respect to manual optimizations, the automatic approach decreases the execution time overhead while only slightly increasing the required code size.