论文部分内容阅读
分析了交换结构规模不同的2类典型的虚通道路由器体系结构,指出了交换结构规模对调度机制和仲裁器体系结构的显著影响.在轮转(round-robin)仲裁器PPE的基础上,提出了分别适用于上述2类路由器的改进型PPE仲裁器体系结构.改进型仲裁器将原有关键路径中的处理步骤转移到非关键路径中,并行实现了关键路径中的重要处理步骤,通过缩短关键路径改善了仲裁器的性能;同时通过调整仲裁流程消除了上述改进对资源成本的负面影响.基于TSMC 0.35μm CMOS工艺实现了不同规模的PPE和2种改进型仲裁器体系结构.实验结果表明上述2种改进型仲裁器比PPE仲裁器的性能分别提高12%和17%以上,并且在多数情况下同时减少了实现面积.
Two kinds of typical virtual channel router architectures with different switch fabric sizes are analyzed, and the significant impact of switch fabric size on scheduling mechanism and arbiter architecture is pointed out.On the basis of round-robin arbiter PPE, Which are respectively applied to the improved PPE arbiter architecture of the above two types of routers.An improved arbiter transfers the processing steps in the original critical path to the non-critical path and implements in parallel important processing steps in the key path, and by shortening the key Path to improve the performance of the arbiter.At the same time, by adjusting the arbitration process, the negative impact of the above improvements on the resource cost is eliminated.On the basis of TSMC 0.35μm CMOS process, PPE of different scales and two kinds of improved arbiter architectures are realized.The experimental results show that The performance of the two improved arbiters increased by 12% and 17%, respectively, over PPE arbiters, and in most cases reduced their implementation area.