Measurement-Based Timing Analysis for Reconfigurable Embedded Systems

Abstract

Reconfigurable embedded computing opens many interesting possibilities for efficiency and reliability improvement. Still, it is necessary to verify whether the reconfigurable embedded computer system fulfills its timing requirements. Purely static worst-case execution time (WCET) analysis is not the best choice for verifying the timing behavior of reconfigurable embedded computer systems, as frequent re-modeling of different hardware configurations is not practicable. We describe measurement-based timing analysis as an approach that is better applicable for reconfigurable applications. Measurement-based timing analysis is a hybrid approach that combines static program analysis techniques, borrowed from static WCET analysis, with empirical learning of the application’s timing behavior by performing systematic measurements. But even the MBTA approach is challenged by the reconfigurable computing paradigm by the need for adequate coverage of the reconfigurable fabric.