Abstract
An algorithm for pipelining loop execution in the presence of loop carried dependences is presented. Both the initiation interval and the turnaround time of a schedule are optimized. Given constraints on the number of functional units and buses, an initiation interval is determined, and then the operations are incrementally partitioned into blocks to fit into the execution windows. A refinement procedure is incorporated to improve the turnaround time. This approach differs from others in that the scheduled operations are iteratively moved up and down to accommodate the ready, yet unscheduled, operations.