论文部分内容阅读
对于传统的平面结构,三维片上网络(3D NoC)具有更好的集成度和性能,在单芯片内部可以集成更多的处理器核。3D NoC作为2D NoC的结构拓展,在性能提高和低功耗设计方面更具优越性,成为多核系统芯片结构的主流架构。映射就是应用某种算法寻找一种最优方案,将通信任务图的子任务分配到NoC的资源节点上,保证NoC的通信能耗最小。参照2D NoC的研究方法,提出了针对3D网格NoC的通信能耗模型,采用蚁群算法实现了面向通信能耗的NoC映射。实验结果表明,面向不同网络规模的3D网格NoC平台,蚁群映射同随机映射相比,通信能耗降低可以达23%～42%。
For the traditional planar structure, 3D NoC has better integration and performance, and can integrate more processor cores in a single chip. As a 2D NoC architecture, 3D NoC is superior in performance improvement and low power consumption design, and becomes the mainstream architecture of multi-core system chip architecture. Mapping is to use an algorithm to find an optimal solution. The subtasks of the communication task graph are allocated to the NoC resource nodes to ensure that NoC consumes the least energy. Referring to the research method of 2D NoC, a communication energy consumption model for 3D grid NoC is proposed, and the NoC mapping for communication energy consumption is realized by using ant colony algorithm. Experimental results show that compared with random mapping, ant colony mapping can reduce communication energy consumption by up to 23% -42% for 3D grid NoC platforms with different network sizes.