Partial expansion of dataflow graphs for resource-aware scheduling of multicore signal processing systems

Abstract

The complex design spaces associated with state-of-the-art, multicore signal processing systems pose significant challenges in realizing designs with high productivity and quality. The Partial Expansion Graph (PEG) implementation model was developed to help address these challenges by enabling more efficient exploration of the scheduling design space for multicore digital signal processors. The PEG allows designers and design tools to systematically adjust and adapt the amount of parallelism exposed from applications depending on the targeted platform. In this paper, we develop new algorithms for scheduling and mapping systems implemented using PEGs. Collectively, these algorithms operate in three steps. First, the amount of data parallelism in the application graph is tuned systematically over many iterations to profit from the available cores in the target platform. Then a mapping algorithm that uses graph analysis is developed to distribute data and task parallel instances over different cores while trying to balance the load of all processing units to make use of pipeline parallelism. Finally, we use a novel technique for performance evaluation by implementing the scheduler and a customizable solution on the programmable platform. We demonstrate the utility of our PEG-based scheduling and mapping algorithms through experiments on real application models and various synthetic graphs.