基于免疫遗传算法的NoC映射研究-测试计量技术及仪器专业论文.docxVIP

摘 摘 要 Abstra Abstract 万方数据 万方数据 - - PAGE IV - 万方数据 摘 要 片上网络是集成电路发展的必由之路，是在深亚微米技术发展下的必然趋势。 NoC 的设计关系到很多方面，而处理单元映射问题在其设计过程中起着重要的作用。 但是，NoC 映射问题是一个 NP 难题。从映射算法着手研究，不断地改进算法本身的 缺陷，加强算法与映射应用的联系，同时构建更加合适的数学模型。只有通过不断地 探索、研究、改进，才能得到更好的最优解。 本文从改进遗传算法本身缺陷这一角度出发，引入免疫遗传算法，构建了抗体选 择函数。同时简要的分析了几种常见拓扑结构的优缺点，进行相关参数比较。在分析 NoC 功耗模型的基础上，将免疫遗传算法应用其中，对通讯功耗进行优化。深入探 讨算法中的各参数的作用，并研究各参数的设置对优化映射结果的影响。 采用 TGFF 软件包产生通讯任务图作为对象，在二维网络拓扑结构下做了以下实 验：首先验证免疫遗传算法有很好的优化效果，然后将每一代最优个体消耗的功耗值 用 matlab 进行画图，说明算法能够很好的收敛；其次验证算法相关参数在算法优化 过程中起到一定的作用和影响；最后和文献中的遗传算法优化进行结果对比，证明免 疫遗传算法更具优势。 关键词：片上网络系统；免疫遗传算法；映射优化 - I - Abstract In the development of deep sub-micron technology，network-on-chip is the only way in the trend of integrated circuit. The design of NoC involves many aspects, and the mapping of processing unit is the very important part in the whole design. It decides the system performance is good or bad. As know to all, the mapping is a NP problem, we should start from the algorithm. First of all, we need to improve the algorithm owe defect. Secondly, we have to build a more suitable mathematical model. The last, strengthen the relationship between the algorithm and the mapping model. Only in these ways we can get a better optimal solution. This paper use immune genetic algorithm to optimize the mapping of net work -on- chip and introduce a selection function. This paper also analysis of the advantages and disadvantages of common topologies briefly. And the paper discusses relative parameters of common topologies. Based on the analysis of the model of the power consumption, the paper take IGA applied to communication power optimization in 2D-Mesh network. In the application process, the paper studies of various parameters and parameter setting of the algorithm deeply, and also analysis the impact on the mapping optimization. The paper takes TGFF package as the experimental task graph generation communication objects. Under the 2D-Mesh network topology the following simulation results have done. Firstly it