面向NoC多核芯片组的任务映射算法  

Mapping algorithm focus on the NoC multi-core chipset

作　　者：侯宁[1] 张多利[1] 

机构地区：[1]合肥工业大学微电子设计研究所,合肥230009

出　　处：《电子测量与仪器学报》2011年第8期695-699,共5页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(编号:61006024)资助项目;安徽省自然科学基金(编号:090412029)资助项目

摘　　要：片上网络(Network On Chip,NoC)是一种高效的片上互连技术。由于NoC的数据通讯具有并发、分离的特性,因此可以方便的在板级集成多块NoC多核芯片协同工作,构成NoC多核芯片组,快速提供更强大的处理能力。板级通讯的数据链路带宽远小于芯片内的带宽,因此必须尽力减小芯片间的数据通讯量。针对这一问题,提出一种面向NoC多核芯片组的任务映射算法。实验表明,该算法可以使芯片间的数据通讯量比初始映射降低24%,能显著提高系统性能。The Network-on-Chip （NoC） is a promising technique for the on-chip interconnection. And the data transfer of NoC is unattached and simultaneous, so several NoC multi-core chips which cooperate with each other, called NoC chipset, set up a powerful computing system. Because the bandwidth between chips is very low, so we must reduce this part of traffic. This paper put forward an application mapping algorithm focus on NoC multi-core chipset. Experiments show with this algorithm, we can effectively reduce the traffic between chips, and prove system performance.

关 键 词：片上网络 NoC多核芯片组 映射算法 路由协议 

分 类 号：TN402[电子电信—微电子学与固体电子学]