{"title":"Multi-applications mapping platform based on hardware and software","authors":"Zhichao Xu, Yiou Chen, X. Ling","doi":"10.1145/3318265.3318267","DOIUrl":null,"url":null,"abstract":"The mapping problem of the NoC (Network on Chip) is NP-hard, each mapping algorithm has its disadvantages for different applications. Traditionally, mapping algorithms verification are based on the software, such as Noxim, which is too slow for large scale NoC emulation. For the reason, we propose a NoC mapping platform and an emulation system with virtualization technology based on FPGA. The experiment show that, the platform can select appropriate mapping algorithms for multi applications, and map-ping algorithm can be improved by the feedback emulation result. Using virtualization technology, the Mesh NoC with 72 node consume only 14% hardware resource on xc7k325tf-fg900-2. We implement the SCPU to generate the data characteristic of the application being mapped with only 2% resource overhead.","PeriodicalId":241692,"journal":{"name":"Proceedings of the 3rd International Conference on High Performance Compilation, Computing and Communications","volume":"3 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2019-03-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the 3rd International Conference on High Performance Compilation, Computing and Communications","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3318265.3318267","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
The mapping problem of the NoC (Network on Chip) is NP-hard, each mapping algorithm has its disadvantages for different applications. Traditionally, mapping algorithms verification are based on the software, such as Noxim, which is too slow for large scale NoC emulation. For the reason, we propose a NoC mapping platform and an emulation system with virtualization technology based on FPGA. The experiment show that, the platform can select appropriate mapping algorithms for multi applications, and map-ping algorithm can be improved by the feedback emulation result. Using virtualization technology, the Mesh NoC with 72 node consume only 14% hardware resource on xc7k325tf-fg900-2. We implement the SCPU to generate the data characteristic of the application being mapped with only 2% resource overhead.
片上网络(Network on Chip)的映射问题是NP-hard问题,针对不同的应用,每种映射算法都有其缺点。传统的映射算法验证是基于软件的,如Noxim,这对于大规模的NoC仿真来说太慢了。为此,我们提出了基于FPGA的NoC映射平台和基于虚拟化技术的仿真系统。实验表明,该平台可以针对多种应用选择合适的地图绘制算法,并通过反馈仿真结果对地图绘制算法进行改进。采用虚拟化技术,72节点的Mesh NoC在xc7k325tf-fg900-2上仅消耗14%的硬件资源。我们实现SCPU来生成被映射应用程序的数据特征,只需要2%的资源开销。