Bongjae Kim, Jinmang Jung, Hong Min, Junyoung Heo, Hyedong Jung
{"title":"基于MPI环境的多gpu性能评估","authors":"Bongjae Kim, Jinmang Jung, Hong Min, Junyoung Heo, Hyedong Jung","doi":"10.1145/3129676.3129716","DOIUrl":null,"url":null,"abstract":"GPU-based computations are widely used in various computing areas because GPU provides very high computing performance when compared to typical CPU. In this paper, we evaluate and analyze the computing performance of multiple GPUs based on MPI environments. We examine the performance of sparse matric-vector multiply (SpMV). SpMV is one of the most heavily used components in many scientific applications. Based on the performance evaluation results, generally, the execution time of SpMV is decreased as the number of GPUs increase. In some case, the performance was reduced according to the computation overhead, the memory copy overhead among GPUs, and the characteristics of sparse matrices.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2017-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"3","resultStr":"{\"title\":\"Performance Evaluations of Multiple GPUs based on MPI Environments\",\"authors\":\"Bongjae Kim, Jinmang Jung, Hong Min, Junyoung Heo, Hyedong Jung\",\"doi\":\"10.1145/3129676.3129716\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"GPU-based computations are widely used in various computing areas because GPU provides very high computing performance when compared to typical CPU. In this paper, we evaluate and analyze the computing performance of multiple GPUs based on MPI environments. We examine the performance of sparse matric-vector multiply (SpMV). SpMV is one of the most heavily used components in many scientific applications. Based on the performance evaluation results, generally, the execution time of SpMV is decreased as the number of GPUs increase. In some case, the performance was reduced according to the computation overhead, the memory copy overhead among GPUs, and the characteristics of sparse matrices.\",\"PeriodicalId\":326100,\"journal\":{\"name\":\"Proceedings of the International Conference on Research in Adaptive and Convergent Systems\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2017-09-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"3\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings of the International Conference on Research in Adaptive and Convergent Systems\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1145/3129676.3129716\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3129676.3129716","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Performance Evaluations of Multiple GPUs based on MPI Environments
GPU-based computations are widely used in various computing areas because GPU provides very high computing performance when compared to typical CPU. In this paper, we evaluate and analyze the computing performance of multiple GPUs based on MPI environments. We examine the performance of sparse matric-vector multiply (SpMV). SpMV is one of the most heavily used components in many scientific applications. Based on the performance evaluation results, generally, the execution time of SpMV is decreased as the number of GPUs increase. In some case, the performance was reduced according to the computation overhead, the memory copy overhead among GPUs, and the characteristics of sparse matrices.