基于通信链路的NoC映射算法  

An NoC mapping algorithm based on communication links

作　　者：金世燕[1] 姚远程[1] 秦明伟[1] 

机构地区：[1]西南科技大学信息工程学院特殊环境机器人技术四川省重点实验室,四川绵阳621010

出　　处：《电子技术应用》2016年第8期121-124,共4页Application of Electronic Technique

基　　金：国防基础科研计划项目(B3120133002);西南科技大学创新团队基金(tdtk02);西南科技大学研究生创新基金(15ycx121)

摘　　要：片上网络映射算法对系统通信能耗、延时等性能具有重大影响。基于通信链路通信量大小,提出一种改进的多目标遗传映射算法,以降低系统能耗和延时。算法中提出根据通信链路通信量大小决定任务在网络拓扑中映射位置的方式来产生初始染色体,并通过改进的变异操作产生新的子代,有效降低了算法复杂度,加快了算法的收敛速度。实验通过NIRGAM仿真平台进行,结果表明,与传统的多目标遗传算法相比,在实际应用DVOPD中,能耗降低了49.76%,通信延时降低了53.23%;在VOPD实验中,能耗和延时分别降低了29.54%和32.45%;而MPEG-4的能耗和延时则分别降低了45.72%和49.40%。同样,提出的算法与模拟退火算法相比,能耗和延时性能也有明显提高。Network- on- Chip（ No C） mapping algorithm has significant impact on system power consumption, latency and other performances. An improved multi- objective genetic mapping algorithm based on the communication links is proposed to decrease power consumption and latency. The algorithm designs a new chromosome initialization method whose communication links directly decide the positions of tasks in network. And a new mutation operator is used to getting offspring. The new algorithm reduces the com-plexity effectively and can accelerate the rate of converging. The experiment is implemented on NIRGAM, results show that power consumption and latency on DVOPD decreased 49. 76 % and 53. 23 % respectively. Compared with traditional multi- objective genetic algorithm, the result on VOPD can get power consumption 29. 54 % reduced and latency 32. 45 % reduced respectively. When come to MPEG- 4, power consumption decreased 45. 72 % and latency decreased 49. 40 %. Similarly, compared with simulated annealing algorithm, performances of power consumption and latency of the proposed algorithm improved obviously.

关 键 词：片上网络 映射 多目标遗传算法 通信链路 能耗 延时 

分 类 号：TN915[电子电信—通信与信息系统] TP301[电子电信—信息与通信工程]