Mobility overlap-removal based timing-constrained scheduling

In the High-Level Synthesis (HLS), the scheduling is a key problem. Recently, the max-flow scheduling method is proposed for resource-constrained scheduling based on the mobility overlap removal. In the max-flow scheduling method, it is a hard problem how to remove the overlaps optimally. With an improper mobility overlap removal, there might be no feasible solutions with resource constraints. In this work, the mobility overlap removal-based method is extended to deal with timing-constrained scheduling problem, where the timing constraint is given as an input and the resource usage should be minimized. A simulated-annealing based method is used to explore the optimal mobility overlap removal. And, given the overlap-free mobilities, a convex-cost flow based algorithm is proposed to schedule single-cycle operations optimally. At the same time, greedy algorithms are devised to schedule multi-cycle operations. The experimental results show the effectiveness of the proposed method.