Extending dataflow programs with throughput properties

In the context of multi-core processors and the trend toward many-core, dataflow programming can be used as a solution to the parallelization problem. By decoupling computation from communication, this paradigm naturally exposes parallelism in several ways. In this work we propose language extensions for expressing throughput properties over dataflow programs together with a run-time mechanism for the observation of events meaningful to compute the effective throughput. We show the limited impact of such mechanisms on the application overall performances. We also review existing run-time adaptation mechanisms that may be used in a dataflow context to satisfy throughput requirements.