A Dynamic Programming Approach to Energy-Efficient Scheduling on Multi-FPGA based Partial Runtime Reconfigurable Systems

Abstract

This paper has been studied an important issue of energy-efficient scheduling on multi-FPGA systems. The main challenges are integral allocation, reconfiguration overhead and exclusiveness and energy minimization with deadline constraint. To tackle these challenges, based on the theory of dynamic programming, we have designed and implemented an energy-efficient scheduling on multi-FPGA systems. Differently, we have presented a MLPF algorithm for task placement on FPGAs. Finally, the experimental results have demonstrated that the proposed algorithm can successfully accommodate all tasks without violation of the deadline constraint. Additionally, it gains higher energy reduction 13.3% and 26.3% than that of Particle Swarm Optimization and fully balanced algorithm, respectively.