Hardware-software co-synthesis of low power real-time distributed embedded systems with dynamically reconfigurable FPGAs

In this paper, we present a multi-objective hardware-software co-synthesis system for multi-rate, real-time, low power distributed embedded systems consisting of dynamically reconfigurable FPGAs, processors, and other system resources. We use an evolutionary algorithm based framework for automatically determining the quantity and type of different system resources, and then assigning tasks to different processing elements (PEs) and task communications to communication links. For FPGAs, we propose a two-dimensional, multi-rate cyclic scheduling algorithm, which determines task priorities based on real-time constraints and reconfiguration overhead information, and then schedules tasks based on the resource utilization and reconfiguration condition in both space and time. The FPGA scheduler is integrated in a list-based system scheduler. To the best of our knowledge, this is the first multi-objective co-synthesis system, which uses dynamically reconfigurable devices to synthesize a distributed embedded system, to target simultaneous optimization of system price and power. Experimental results indicate that our method can reduce schedule length by an average of 41.0% and reconfiguration power by an average of 46.0% compared to the previous method. It also yields multiple system architectures which trade off system price and power under real-time constraints.