An Automated Scheduling and Partitioning Algorithm for Scalable Reconfigurable Computing Systems

As reconfigurable computing (RC) platforms are becoming increasingly large-scale and heterogeneous, efficiently scheduling and partitioning applications on these platforms is a growing challenge. While previous approaches support scheduling and HW/SW partitioning for various reconfigurable architectures, none of these approaches have been designed or shown to support large-scale or multi-node RC systems. This paper presents an algorithm that to our knowledge is the first designed to support automated scheduling and HW/SW partitioning on large-scale RC systems. The algorithm presented here uses a two-stage process. The first stage generates an initial schedule using novel list-based scheduling extensions, which feeds an iterative cycle in stage two to search for moves to further optimize the schedule. Initial results show the algorithm can efficiently produce quality schedules for parallel applications on a large-scale RC platform.