Impact of task distribution, processor configurations and dynamic clock frequency scaling on the power consumption of FPGA-based multiprocessors

As only the currently required functionality on a dynamic reconfigurable FPGA-based system is active, a good performance per power ratio can be achieved. To find such a good performance per power ratio for a given application is a difficult task, as it requires not only knowledge of the behavior of the application, but also knowledge of the underlying hardware architecture and its influences on the performance and the static and dynamic power consumption. Is it for example better to use two processors running at half the clock frequency than a single processor? The main contributions of this paper are: the description of a tool flow to measure the power consumption for multiprocessor systems in Xilinx FPGAs, a novel runtime adaptive architecture for analyzing the performance per power tradeoff and for dynamic clock frequency scaling based-on the inter-processor communication. Furthermore, we use three different application scenarios to show the influence of the clock frequency, different processor configurations and different application partitions onto the static and dynamic power consumption as well as onto the overall system performance.