Evaluation of some multiprocessor scheduling techniques of atomic operations for recursive DSP filters

This paper is focused on the comparison and evaluation of optimal and near optimal multiprocessor static scheduling techniques for recursive DSP algorithms. The scheduling problem is modeled by a data flow graph of atomic operations with known computational delay. Two different scheduling approaches are considered. One consists in techniques based on the scheduling of the graph that results from the conversion of the original cyclic graph into an acyclic one. The second approach, based on the analysis of the cycles of the graph, consists of methods that lead to rate optimal solutions