Why Model Checking Can Improve WCET Analysis

Calculating predictions for an upper bound of the execution time of real-time tasks in embedded systems is a necessary step in designing such systems. There exist successful analysis methods, based on abstract interpretation and integer linear programming (ILP) for that problem. In [12] it is stated, that model checking is not adequate for this task. The approach presented in this paper shows that model checking is adequate and, furthermore, can improve the results. This is done by defining an automaton based semantic for control flow graphs of programs for abstract and concrete instruction cache analysis. A binary search based bunch of model checker runs is used to calculate the upper bound of execution time.