NoC处理单元随机舍入的启发式应用映射  

Randomized Rounding Heuristic for Application Mapping to NoC Processing Elements

作　　者：宋朝晖[1,2] 马光胜[1] 宋大雷[3] 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,哈尔滨150001 [2]黑龙江电力职工大学计算机网络中心,哈尔滨150030 [3]中国海洋大学工程学院,青岛266001

出　　处：《计算机辅助设计与图形学学报》2011年第7期1263-1269,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60273081)

摘　　要：在性能约束的前提下,为了降低能量消耗,针对一个集成了异构IP块以分层星型拓扑互连的片上网络(NoC),采用多电压电平操作,运用一种统一方法来解决应用映射问题,并用混合整数线性程序公式化问题,提出了基于随机舍入的NoC处理单元启发式高效应用映射——HAMU法.实验结果表明,HAMU法的效能明显优于其他启发式方法.With the performance constraints,a unified approach is presented to reduce energy and solve the application-mapping problem on a heterogeneous network-on-chip（NoC） platform,which incorporates heterogeneous intellectual properties（IPs）,is interconnected in a hierarchical star topology and operating at multiple voltage levels.Mixed Integer Linear Program is then used to formulate the problem.Finally,a randomized rounding based heuristic algorithm（HAMU） is proposed for energy efficient application mapping on NoC processing elements.Experimental results demonstrated that the HAMU approach is more efficient for application-mapping comparing with other heuristic algorithms.

关 键 词：片上网络 混合整数线性程序 随机舍入 启发式 应用映射 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]