Scheduling of DSP data flow graphs onto multiprocessors for maximum throughput

A novel technique to obtain a rate- and processor-optimal schedule for a fully-static data flow graph (DFG) onto a multiprocessor system is presented. In this technique, Floyd-Warshall's shortest path algorithm is used to evaluate the relative firing times of the nodes of the given DFG. Despite its implementation simplicity, the proposed technique has a lower time complexity than all the previously proposed techniques. The technique is tested on various benchmark problems to demonstrate its optimal performance.