Evolutionary algorithms for the verification of execution time bounds for real-time software

Real-time systems must produce results according to a predefined time schedule. Therefore, their operation speed is critical. Timing analysis is crucial for the verification of a program's timing behaviour and an important part of design, testing and assessment. It measures how well a system matches its specifications. Static timing analysis concentrates on the evaluation of a program's internal structure to predict execution time bounds. But it does not consider a program's input parameters which are mainly responsible for the dynamic behaviour. Dynamic timing analysis can produce the most accurate assessment of run-time behaviour through the analysis of the interactions of a module's input parameters, but it is considered to be impractical because of the possible combinatorial explosion of the parameter combination space. However, powerful searching strategies such as evolutionary algorithms now make dynamic timing analysis of systems feasible. They only need to search through a fraction of the input parameters' total combination space and can be used to replace or supplement static techniques. Experiments with the new methodology have already shown significant improvement in the prediction of timing constraints, although the technique is merely in an experimental state.