Direct handling of infeasible paths in the event dependency analysis

摘要

While common task models in real-time analysis assume a task as being activated by a single event and producing a single outgoing event after its full completion, the event dependency analysis extended the model to allow for multiple outgoing events to occur already during the run-time of a single job. It uses the structure of a timing-annotated control flow graph to calculate the maximal densities of outgoing events. Previous work has shown that knowledge about infeasible paths in the control flow graph can be used to relax the event densities by modifying the graph prior to performing the event dependency analysis. However, that approach leads to a nearly explicit path enumeration. In this paper, we show how the event dependency analysis can be modified to directly exclude infeasible paths in order to relax the densities of outgoing events.