Kalindu Herath, Alok Prakash, Udaree Kanewala, T. Srikanthan
{"title":"Communication-Aware Module Placement for Power Reduction in Large FPGA Designs","authors":"Kalindu Herath, Alok Prakash, Udaree Kanewala, T. Srikanthan","doi":"10.1109/ISVLSI.2018.00047","DOIUrl":null,"url":null,"abstract":"Modern multi-million logic FPGAs allow hardware designers to map increasingly large designs into FPGAs. However, traditional FPGA CAD flows scale poorly for large designs, often producing low quality solutions in terms of performance and power in such cases. To improve the design productivity, modular design methodology partitions a large design into subsystems, compiles them individually and finally collates the individual solutions to complete the mapping process. Existing work has attempted to partition large designs into smaller subsystems, based on the intra-subsystem communication frequencies, to reduce routing power dissipation. However, inter-subsystem communication has not been considered, especially, during the placement stage. In this paper, we first show the adverse effect of ignoring the inter-subsystem communication during the placement stage. Next, we propose an inter-subsystem communication-aware placement technique using a Simulated Annealing based approach to achieve significant power savings. Experimental results show over 7% reduction in routing power when compared to the existing state-of-the-art partitioning flow that does not consider inter-subsystem communication, while the routing power reduction is over 11% when compared to a commercial CAD tool such as Altera Quartus.","PeriodicalId":114330,"journal":{"name":"2018 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2018-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2018 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ISVLSI.2018.00047","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
Abstract
Modern multi-million logic FPGAs allow hardware designers to map increasingly large designs into FPGAs. However, traditional FPGA CAD flows scale poorly for large designs, often producing low quality solutions in terms of performance and power in such cases. To improve the design productivity, modular design methodology partitions a large design into subsystems, compiles them individually and finally collates the individual solutions to complete the mapping process. Existing work has attempted to partition large designs into smaller subsystems, based on the intra-subsystem communication frequencies, to reduce routing power dissipation. However, inter-subsystem communication has not been considered, especially, during the placement stage. In this paper, we first show the adverse effect of ignoring the inter-subsystem communication during the placement stage. Next, we propose an inter-subsystem communication-aware placement technique using a Simulated Annealing based approach to achieve significant power savings. Experimental results show over 7% reduction in routing power when compared to the existing state-of-the-art partitioning flow that does not consider inter-subsystem communication, while the routing power reduction is over 11% when compared to a commercial CAD tool such as Altera Quartus.