Tight temporal bounds for dataflow applications mapped onto shared resources

We present an analysis method that provides tight temporal bounds for applications modeled by Synchronous Dataflow Graphs and mapped to shared resources. We consider the resource sharing effects on the temporal behaviour of the application by embedding worst case resource availability curves in the symbolic simulation of the application graph. Symbolic simulation of the application results in a (max, +) characterization matrix. This matrix specifies a set of recursive linear equations in (max, +) algebra that bound the worst case execution of the application. We obtain tighter temporal bounds on the completion times of tasks than state of the art analysis. This is achieved by improving the response times of the tasks by identifying possible consecutive task executions on the resources. This enables us to use accumulated response times which are less pessimistic. Applying the new approach to real-life applications gives significant improvements over the bounds compared to state of the art.