H. Kim, John Kim, Woong Seo, Yeon-Gon Cho, Soojung Ryu
{"title":"在gpgpu中提供高性价比的片上网络带宽","authors":"H. Kim, John Kim, Woong Seo, Yeon-Gon Cho, Soojung Ryu","doi":"10.1109/ICCD.2012.6378671","DOIUrl":null,"url":null,"abstract":"Network-on-chip (NoC) bandwidth has a significant impact on overall performance in throughput-oriented processors such as GPG-PUs. Although it has been commonly assumed that high NoC bandwidth can be provided through abundant on-chip wires, we show that increasing NoC router frequency results in a more cost-effective NoC. However, router arbitration critical path can limit the NoC router frequency. Thus, we propose a direct all-to-all network overlaid on mesh (DA2mesh) NoC architecture that exploits the traffic characteristics of GPGPU and removes arbitration from the router pipeline. DA2mesh simplifies the router pipeline with 36% improvement of performance while reducing NoC energy by 15%.","PeriodicalId":313428,"journal":{"name":"2012 IEEE 30th International Conference on Computer Design (ICCD)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2012-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"35","resultStr":"{\"title\":\"Providing cost-effective on-chip network bandwidth in GPGPUs\",\"authors\":\"H. Kim, John Kim, Woong Seo, Yeon-Gon Cho, Soojung Ryu\",\"doi\":\"10.1109/ICCD.2012.6378671\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Network-on-chip (NoC) bandwidth has a significant impact on overall performance in throughput-oriented processors such as GPG-PUs. Although it has been commonly assumed that high NoC bandwidth can be provided through abundant on-chip wires, we show that increasing NoC router frequency results in a more cost-effective NoC. However, router arbitration critical path can limit the NoC router frequency. Thus, we propose a direct all-to-all network overlaid on mesh (DA2mesh) NoC architecture that exploits the traffic characteristics of GPGPU and removes arbitration from the router pipeline. DA2mesh simplifies the router pipeline with 36% improvement of performance while reducing NoC energy by 15%.\",\"PeriodicalId\":313428,\"journal\":{\"name\":\"2012 IEEE 30th International Conference on Computer Design (ICCD)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2012-09-30\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"35\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2012 IEEE 30th International Conference on Computer Design (ICCD)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICCD.2012.6378671\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2012 IEEE 30th International Conference on Computer Design (ICCD)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICCD.2012.6378671","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Providing cost-effective on-chip network bandwidth in GPGPUs
Network-on-chip (NoC) bandwidth has a significant impact on overall performance in throughput-oriented processors such as GPG-PUs. Although it has been commonly assumed that high NoC bandwidth can be provided through abundant on-chip wires, we show that increasing NoC router frequency results in a more cost-effective NoC. However, router arbitration critical path can limit the NoC router frequency. Thus, we propose a direct all-to-all network overlaid on mesh (DA2mesh) NoC architecture that exploits the traffic characteristics of GPGPU and removes arbitration from the router pipeline. DA2mesh simplifies the router pipeline with 36% improvement of performance while reducing NoC energy by 15%.