{"title":"A Multi-node MPI Implementation on the CellBE Processor","authors":"Chao Liu, Xingjun Zhang, Jinghua Feng, Xiaoshe Dong","doi":"10.1109/ChinaGrid.2010.41","DOIUrl":null,"url":null,"abstract":"Due to the heterogeneous multi-core architecture of the CellBE processor, the existing MPI implementation cannot effectively exert the SPE computing capability. In this paper, we design and implement a multi-node MPI programming and running environment for CellBE processor. The MPI communication library prototype, which mainly includes the basic point-to-point communication and the multicast communication operations, is implemented in this cell-MPI model. By this work, the programmers can exploit the SPE’s computing capability to improve the program running efficiency using the traditional MPI programming model; the burden of porting application to the CellBE processor is much reduced. The experimental results show that the presented MPI programming model not only exerts the SPE’s computing capability but also keeps the efficient MPI communication performance.","PeriodicalId":429657,"journal":{"name":"2010 Fifth Annual ChinaGrid Conference","volume":"12 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2010-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2010 Fifth Annual ChinaGrid Conference","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ChinaGrid.2010.41","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
Due to the heterogeneous multi-core architecture of the CellBE processor, the existing MPI implementation cannot effectively exert the SPE computing capability. In this paper, we design and implement a multi-node MPI programming and running environment for CellBE processor. The MPI communication library prototype, which mainly includes the basic point-to-point communication and the multicast communication operations, is implemented in this cell-MPI model. By this work, the programmers can exploit the SPE’s computing capability to improve the program running efficiency using the traditional MPI programming model; the burden of porting application to the CellBE processor is much reduced. The experimental results show that the presented MPI programming model not only exerts the SPE’s computing capability but also keeps the efficient MPI communication performance.