面向特定应用的片上网络低能耗拓扑生成方法  

Low energy topology generation approach for application-specific network on chip

作　　者：葛芬[1] 吴宁[1] 

机构地区：[1]南京航空航天大学信息科学与技术学院,江苏南京210016

出　　处：《系统工程与电子技术》2010年第8期1754-1759,共6页Systems Engineering and Electronics

基　　金：江苏省自然科学基金(BK2008387);江苏省研究生创新基金(CX07B_105z)资助课题

摘　　要：为解决面向特定应用的片上网络(network on chip,NoC)设计中的低能耗问题,提出基于遗传算法的层次化拓扑生成方法(genetic algorithm based topology generation,GATG)。该方法以降低网络通信能耗为目标,在满足带宽和延时约束下,根据给定应用的通信需求和路由器结构特征,自动将IP核映射到所选路由器单元。同时提出一种基于递归实现的路径构建算法嵌入GATG,用于确定IP核间的通信路径,从而构建路由器之间的链路,最终生成专用的不规则的NoC拓扑结构。实验结果表明,GATG生成结果与采用规则二维Mesh结构相比平均降低能耗46.5%,并可节省52.7%的路由器资源开销。A genetic algorithm based hierarchical topology generation approach（GATG） is proposed to solve the problem of low energy design of application-specific network on chip（NoC）.The aim is to reduce the network communication energy.Under the constraints of the bandwidth and latency,GATG automatically maps IP cores onto the selected routers according to the communication requirements of given application and the characteristics of router architectures.In addition,a recursion based routing path construction algorithm embedded in GATG is proposed to construct links between routers,and finally the application-specific irregular NoC topology is formed.Experimental results show that GATG achieves a reduction by an average of 46.5% in energy consumption and 52.7% in router resource overhead compared with regular 2D Mesh topology.

关 键 词：片上网络 拓扑生成 映射 路由 低能耗 遗传算法 

分 类 号：TN47[电子电信—微电子学与固体电子学]