Power reduction through specific instruction scheduling based on Hardware/Software Co-Design

Abstract

In this paper, an instruction-level power reduction model for the low power system-on-a-chip is proposed, which combines the hardware and software design together. Firstly, to reduce the power consumption via hardware design, this model is equipped with a specific instruction extraction process, which utilizes a sub-graph matching algorithm. Then a scheduling algorithm is integrated in this model to achieve power compression via reducing the memory access number. Finally, a set of Fir filter programs are power-driven optimized using the proposed model based on hardware/software co-design strategy, and the experimental results indicate that this model can reduce the power consumption effectively.