缓解异构MPSoC电迁移效应的任务调度算法

发布时间：2018-10-07 19:42

【摘要】：随着集成电路制造工艺进入到纳米时代,日益严重的电路老化给多处理器片上系统(MPSo C)可靠度带来严峻挑战.针对在性能异构MPSo C中,已有的可靠度优化方法没有考虑处理器之间的可靠度差异的问题,提出一种减小处理器可靠度差异的任务调度算法.该算法结合电迁移效应下单个处理器的平均无故障时间模型分析了制约单个处理器可靠度的因素,得出性能异构MPSo C中处理器可靠度差异模型;基于此差异模型,提出一种交叉分配任务调度算法——cross,减小了处理器间可靠度差异,达到整体优化可靠度的目的.实验结果表明,与已有的均衡受压任务调度算法相比,cross任务调度算法下异构MPSo C中处理器的平均无故障时间变异系数降低了3.71%.
[Abstract]:With the development of IC manufacturing process into the nanometer era, the increasing aging of the circuit brings a severe challenge to the reliability of the multiprocessor system (MPSo C). Aiming at the problem that the existing reliability optimization methods in heterogeneous MPSo C do not take into account the reliability differences among processors, a task scheduling algorithm is proposed to reduce the differences in processor reliability. Combined with the average failure time model of a single processor under electromigration effect, this algorithm analyzes the factors that restrict the reliability of a single processor, and obtains a processor reliability differential model in heterogeneous MPSo C. A cross-assignment task scheduling algorithm, cross. is proposed, which reduces the reliability difference among processors and achieves the purpose of overall optimization of reliability. The experimental results show that compared with the existing equalization and compression task scheduling algorithm, the average non-fault time variation coefficient of the processor in heterogeneous MPSo C is 3.71% lower than that of the crossover task scheduling algorithm.
【作者单位】：合肥工业大学计算机与信息学院;合肥工业大学电子科学与应用物理学院;华北电力大学电气与电子工程学院;
【基金】：国家自然科学基金(61274036,61371025,61204027)
【分类号】：TN47;TP332

【参考文献】