Using Software Metrics to Estimate the Impact of Maintenance in the Performance of Embedded Software

This paper proposes a strategy to assist the designer in evaluating the impact of a design choice with respect to the non-functional requirements in embedded systems. We use several regression models to predict physical metrics from design metrics in order to estimate the impact on performance of software changes in the early stages of its development. This prediction can be used both during maintenance and during the first design to compare alternative module decompositions or design changes before implementation. Such an early estimation allows an efficient design space exploration with no penalty in the development time, which are crucial aspects for an embedded system.