面向实时复杂系统的基于片上网络多核平台的映射技术研究  

Energy and performance-aware mapping for real-time complex system based on NoC platform

作　　者：陈亦欧[1] 胡剑浩[1] 陈庚生[1] 

机构地区：[1]电子科技大学通信抗干扰国家级重点实验室,成都611731

出　　处：《计算机应用研究》2012年第7期2589-2592,2628,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(60873076);博士点新教师基金资助项目(200806141015)

摘　　要：为了提高通信与信号处理系统向片上网络(NoC)平台映射的性能,提出一种面向系统能耗和响应时间的新型NoC映射模型。该模型改进了传统映射算法中能耗的计算方法,同时考虑交换网络的能耗和处理节点的能耗,采用多对多映射技术,在带宽约束的基础上,结合复杂通信系统特有的注入率、迭代边界等约束,同时优化能耗和响应时间,提高了映射算法的实用性。仿真实例证明,该映射模型得到的优化方案能降低约15%的系统能耗,能够获得综合的最优性能。This paper proposed a new network on chip（NoC） mapping algorithm which considered system energy consumption and response time to meet the real-time requirements of communication and signal processing system.It proposed an optimized energy model,which contained both energy of interconnection network and energy of processing elements and adopted a multiple-to-multiple mapping scheme.Under the bandwidth constrains and several characteristic constraints of complex communication system,such as iteration bound and injection rate,this model optimized energy consumption and response time simultaneously.The simulation results prove that the proposed mapping algorithm can achieve about 15% energy consumption savings,and has smaller response time than random mapping.

关 键 词：片上网络映射 数字信号处理 能耗 响应时间 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]