Maximum-throughput scheduling with limited resources for iterative data-flow graphs by means of the scheduling-range chart

An algorithm based on an alternative scheduling approach for iterative acyclic and cyclid DFGs (data-flow graphs) with limited resources that exploits inter- and intra-iteration parallelism is presented. The method is based on guiding the scheduling algorithm with the information supplied by a scheduling-range chart. This scheduling range is relative to a reference operation and can be finite or infinite. The information in the scheduling-range chart is used during scheduling in order to optimize the sampling period. For cases where the precedence constraints do not allow a schedule in the originally selected optimal sampling period, the algorithm provides an adjustment procedure, thus always guaranteeing a solution. The delay of the processor-assignment phase increases the efficiency of the algorithm when the operations have processing times different from the unity and are nonpreemptive