Statistical energy optimization on voltage-frequency island based MPSoCs in the presence of process variations

Energy efficiency has become a primary design concern for embedded multiprocessor system-on-chips (MPSoCs). Recently, voltage-frequency island (VFI)-based design paradigm was proposed to optimize system energy by combining with task scheduling. However, the ever-increasing variations cause large uncertainty for delay and power of the processors, resulting in performance parameters of the VFIs also manifesting as statistical distribution. As a result, it is more difficult for the deterministic energy optimization to achieve desirable performance yield, defined as the probability of the design meeting timing constraints of the system. In this paper, we propose a variation-aware statistical energy optimization framework which takes account of performance yield constraints in the overall optimization flow. We observed that statistical features of parameter distributions in homogeneous and heterogeneous platforms are different. We hence define two effective metrics, namely energy optimization sensitivity (EOS) and lowest operating voltage (LOV), to make our framework adapt to both of the platforms. Experimental results demonstrate that on average our framework achieves performance yield improvement of 45% than the deterministic scheme. In respect of energy optimization, on average our framework achieves energy reduction of 33% than the existing statistical task scheduling algorithm.