{"title":"基于远程存储器访问的多尺度MD-FE耦合中的并行尺度传输","authors":"Dorian Krause, R. Krause","doi":"10.1109/eScienceW.2011.8","DOIUrl":null,"url":null,"abstract":"We present a novel remote memory access based approach to the parallelization of a coupled Molecular Dynamics-Finite Element (MD/FE) code and its implementation. Such multiscale simulation codes can be used for, e.g., the simulation of localized nonlinear phenomena as fracture in solids. Caused by the dynamic data distribution of MD codes this requires non-standard communication mechanisms. We investigate different implementation options for realizing the scale transfer between MD and FE in parallel. We present benchmark results obtained on three different parallel architectures which give interesting insights into the performance of our new approach. Additionally, we study the scaling of a full coupled code. To the best of our knowledge, these results are the first available scaling results for this kind of heterogeneous code.","PeriodicalId":267737,"journal":{"name":"2011 IEEE Seventh International Conference on e-Science Workshops","volume":"78 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2011-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Parallel Scale-Transfer in Multiscale MD-FE Coupling Using Remote Memory Access\",\"authors\":\"Dorian Krause, R. Krause\",\"doi\":\"10.1109/eScienceW.2011.8\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"We present a novel remote memory access based approach to the parallelization of a coupled Molecular Dynamics-Finite Element (MD/FE) code and its implementation. Such multiscale simulation codes can be used for, e.g., the simulation of localized nonlinear phenomena as fracture in solids. Caused by the dynamic data distribution of MD codes this requires non-standard communication mechanisms. We investigate different implementation options for realizing the scale transfer between MD and FE in parallel. We present benchmark results obtained on three different parallel architectures which give interesting insights into the performance of our new approach. Additionally, we study the scaling of a full coupled code. To the best of our knowledge, these results are the first available scaling results for this kind of heterogeneous code.\",\"PeriodicalId\":267737,\"journal\":{\"name\":\"2011 IEEE Seventh International Conference on e-Science Workshops\",\"volume\":\"78 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2011-12-05\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2011 IEEE Seventh International Conference on e-Science Workshops\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/eScienceW.2011.8\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2011 IEEE Seventh International Conference on e-Science Workshops","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/eScienceW.2011.8","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Parallel Scale-Transfer in Multiscale MD-FE Coupling Using Remote Memory Access
We present a novel remote memory access based approach to the parallelization of a coupled Molecular Dynamics-Finite Element (MD/FE) code and its implementation. Such multiscale simulation codes can be used for, e.g., the simulation of localized nonlinear phenomena as fracture in solids. Caused by the dynamic data distribution of MD codes this requires non-standard communication mechanisms. We investigate different implementation options for realizing the scale transfer between MD and FE in parallel. We present benchmark results obtained on three different parallel architectures which give interesting insights into the performance of our new approach. Additionally, we study the scaling of a full coupled code. To the best of our knowledge, these results are the first available scaling results for this kind of heterogeneous code.