GPU/CPU异构系统任务节能调度方法仿真

摘要

研究GPU/CPU异构系统任务调度的节能问题.与传统同构体系结构相比,异构系统任务调度呈现较大的随机性和不定性,GPU/CPU异构系统中时间间隙片段呈现了较大的随机性,导致传统调度方法很难建立规则的描述时间片段的模型,调度能耗较高.为解决上述问题,提出了一种改进功耗优化的GPU/CPU异构环境下的任务调度算法,将任务关系图按照依赖关系计算量拆分,并分配到计算节点.在计算节点内根据权重法的思想,统计所有计算节点的处理情况,进而将节点内的子任务调度到合适的处理器.实验结果表明,在不影响应用性能的前提下,降低了异构系统的能耗开销,优化效果明显.%Research the energy consumption optimization of GPU/CPU scheduling system.Task scheduling presents great randomness and uncertainty,and it is difficult to establish a regular describe time slice model,resulting in high energy consumption scheduling.The paper proposed an improved power optimization scheduling algorithm based on GPU/CPU heterogeneous environment.The task scheduling scheme splits the task diagram and distributes to the calculation node according to the dependence relationship calculation.According to the thought of the weighting method,the progress of the statistics of all the compute nodes was analysed.The subtasks of node were scheduled to the appropriate processor.The simulation results show that the algorithm reduces the energy consumption of heterogeneous system.