A novel technique for pipelined scheduling and allocation of data-flow graphs based on genetic algorithms

Genetic algorithms are exploited and applied to the development of a novel technique for the high-level synthesis of digital signal processing algorithms. The proposed technique performs the scheduling and allocation of functional units for the synthesis of both pipelined and non-pipelined data-paths. The salient feature of this technique is that it employs the order crossover operator in combination with a schedule building heuristic to overcome the well known encoding problem encountered when applying genetic algorithms to scheduling problems. The technique is demonstrated through its application to a benchmark fifth-order elliptic wave-digital filter. The results show that the technique produces the optimum schedules for non-pipelined data-paths, but requires further refinements to match the best existing schedules for pipelined data-paths