多任务下I/O设备的动态功耗管理  

Dynamic Power Management for I/O Devices Under Multi-task Environment

作　　者：戚隆宁[1] 张哲[1] 黄少珉[1] 

机构地区：[1]东南大学国家专用集成电路系统工程技术研究中心,南京210096

出　　处：《中国工程科学》2008年第2期60-65,共6页Strategic Study of CAE

摘　　要：减少I/O设备功耗已越来越被嵌入式系统设计者所关注。传统动态功耗管理(DPM)策略在实际的多任务环境下无法得到预期的节能效果。提出基于堆栈的预测性超时(SBPT)策略。该策略通过分析任务的调用和堆栈信息来预测任务对I/O设备的访问模式,并采用多请求源(MSR)模型进行多任务的联合预测。然后根据预测结果分组统计,采用超时技术决策。基于实际负载的仿真实验表明SBPT策略能够适应多任务的应用环境,更稳定更有效地降低了功耗。More embedded system designers pay attention to how to reduce the power consumption of I/O devices. Traditional dynamic power management （DPM） policies only focus on the device requests, and neglect the application features behind the workload. Because of the assumption about the stationary workload, traditional DPM policies can not reach their expected goal under the multi-task environment. The paper presents a stack-based predictive timeout strategy （SBPT）. It can predict the access pattern of the device I/O operations by analyzing the calling and stack information of tasks and combine predictions of multiple tasks to form the global prediction according to the multiple- service-requester model. At last, classify the I/O request by the global prediction and then make the decision with the timeout technique based on the distribution of the grouped requests. An evaluation study of SBFF using the trace-driven simulation is performed. The results show that SBPT can adapt the non-stationary multi-task environment and reduces power consumption more efficiently than other policies.

关 键 词：动态功耗管理 预测性超时 堆栈 多请求源 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]