{"title":"Hy-BTree:一种高效的基于树的拓扑结构,用于基于FPGA的NoC实现","authors":"B. M. Prabhu Prasad, Khyamling Parane, B. Talawar","doi":"10.1109/CONECCT52877.2021.9622580","DOIUrl":null,"url":null,"abstract":"Due to their hierarchical structure, Binary Tree (BTree) topology can be employed in Network-on-Chip (NoC) applications. Because of its lower bisection bandwidth, the performance degradation is observed in communication intensive applications. The Fat tree topology has been proposed to overcome the disadvantages of the BTree topology. But, the complexity of the Fat Tree topology's router becomes more complicated as we move towards the root node of the tree and occupying a huge amount of hardware resources compared to the BTree variant. Instead of going for Fat Tree topology, the number of hops taken by a packet in the BTree topology can be reduced by introducing new links in the network with an increase in the bisection bandwidth. In this work, we propose a variant of BTree topology called Hy-BTree by introducing additional links at the intermediate levels of the network to reduce the number of hops taken for the communication. The proposed design is implemented on the FPGA and compared with the other topologies from state-of-the-art the FPGA based NoC architectures. A reduction in average latency and an improvement in throughput have been observed in Hy-BTree with respect to the BTree network with negligible overhead.","PeriodicalId":164499,"journal":{"name":"2021 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Hy-BTree: An efficient Tree based topology for FPGA based NoC implementation\",\"authors\":\"B. M. Prabhu Prasad, Khyamling Parane, B. Talawar\",\"doi\":\"10.1109/CONECCT52877.2021.9622580\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Due to their hierarchical structure, Binary Tree (BTree) topology can be employed in Network-on-Chip (NoC) applications. Because of its lower bisection bandwidth, the performance degradation is observed in communication intensive applications. The Fat tree topology has been proposed to overcome the disadvantages of the BTree topology. But, the complexity of the Fat Tree topology's router becomes more complicated as we move towards the root node of the tree and occupying a huge amount of hardware resources compared to the BTree variant. Instead of going for Fat Tree topology, the number of hops taken by a packet in the BTree topology can be reduced by introducing new links in the network with an increase in the bisection bandwidth. In this work, we propose a variant of BTree topology called Hy-BTree by introducing additional links at the intermediate levels of the network to reduce the number of hops taken for the communication. The proposed design is implemented on the FPGA and compared with the other topologies from state-of-the-art the FPGA based NoC architectures. A reduction in average latency and an improvement in throughput have been observed in Hy-BTree with respect to the BTree network with negligible overhead.\",\"PeriodicalId\":164499,\"journal\":{\"name\":\"2021 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT)\",\"volume\":\"27 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-07-09\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2021 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/CONECCT52877.2021.9622580\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/CONECCT52877.2021.9622580","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1
摘要
二叉树(BTree)拓扑结构具有层次结构,可用于片上网络(NoC)应用。由于其较低的对分带宽,在通信密集型应用中会出现性能下降。胖树拓扑是为了克服b树拓扑的缺点而提出的。但是,当我们向树的根节点移动时,胖树拓扑的路由器的复杂性变得更加复杂,并且与b树变体相比,占用了大量的硬件资源。在b树拓扑中,数据包的跳数可以通过在网络中引入新的链路并增加平分带宽来减少,而不是采用胖树拓扑。在这项工作中,我们提出了b树拓扑的一种变体,称为hy - b树,通过在网络的中间级别引入额外的链路来减少通信所需的跳数。提出的设计在FPGA上实现,并与基于FPGA的NoC架构的其他拓扑进行了比较。相对于开销可以忽略不计的b - tree网络,hy - b - tree中平均延迟的减少和吞吐量的提高已经被观察到。
Hy-BTree: An efficient Tree based topology for FPGA based NoC implementation
Due to their hierarchical structure, Binary Tree (BTree) topology can be employed in Network-on-Chip (NoC) applications. Because of its lower bisection bandwidth, the performance degradation is observed in communication intensive applications. The Fat tree topology has been proposed to overcome the disadvantages of the BTree topology. But, the complexity of the Fat Tree topology's router becomes more complicated as we move towards the root node of the tree and occupying a huge amount of hardware resources compared to the BTree variant. Instead of going for Fat Tree topology, the number of hops taken by a packet in the BTree topology can be reduced by introducing new links in the network with an increase in the bisection bandwidth. In this work, we propose a variant of BTree topology called Hy-BTree by introducing additional links at the intermediate levels of the network to reduce the number of hops taken for the communication. The proposed design is implemented on the FPGA and compared with the other topologies from state-of-the-art the FPGA based NoC architectures. A reduction in average latency and an improvement in throughput have been observed in Hy-BTree with respect to the BTree network with negligible overhead.