Multi-objective Placement of Reconfigurable Hardware Tasks in Real-Time System

Hardware task placements in dynamically reconfigurable logic need to satisfy different goals such as high placement efficiency, low fragmentation of the reconfigurable logic, and minimization of routing resources. There have been several placement algorithms proposed for each goal. Nevertheless, the algorithm can only satisfy one goal, which results in poor results in the other goal satisfactions. We propose a novel Multi-Objective Hardware Placement (MOHP) method such that all goals are satisfied and if there are conflicts a good trade off is obtained. MOHP is similar to share-based schedulers that try to adjust resource utilizations among different scheduling methods by varying the processor bandwidth. By applying MOHP to some examples, we find that MOHP approximates the best performance for each assessment criterion.