University of Missouri-St. Louis
In this paper, the authors propose a model to analyze the response-time of parallel real-time tasks. The presented model is based on the fork/join model which is internally used by user-level frameworks to exploit the parallelism provided by the underlying architecture. The model considers tasks with fixed priorities and allows real-time jobs to generate an arbitrary number of parallel threads. Each parallel thread is scheduled at runtime by taking into account the timing properties of the job that spawns it.