Towards Adaptable Control Flow Segmentation for measurement-Based Execution Time Analysis

Abstract

During the design of embedded real-time systems, engineers have to consider the temporal behavior of software running on a particular hardware platform. Measurement-based timing analysis is a technique that combines elements from static code analysis with execution time measurements on real physical hardware. Because performing exhaustive measurement is generally not tractable, some kind of abstraction must be used to deal with the combinatoric complexity of real software. We propose an adaptable measurement-based analysis approach that uses the novel flexible abstraction of a segment graph to model control flow at varying levels of detail. We also present preliminary experimental results produced by a prototype implementation.