{"links":{},"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00192841","sets":["1164:2240:9411:9646"]},"path":["9646"],"owner":"44499","recid":"192841","title":["AI橋渡しクラウドABCIにおけるLinpack benchmarkの最適化と性能評価"],"pubdate":{"attribute_name":"公開日","attribute_value":"2018-12-10"},"_buckets":{"deposit":"b3855d6f-0b93-43c5-a9b4-efcaeea758dc"},"_deposit":{"id":"192841","pid":{"type":"depid","value":"192841","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"AI橋渡しクラウドABCIにおけるLinpack benchmarkの最適化と性能評価","author_link":["451027","451024","451029","451022","451028","451021","451026","451031","451020","451030","451018","451023","451025","451019"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"AI橋渡しクラウドABCIにおけるLinpack benchmarkの最適化と性能評価"},{"subitem_title":"Performance evaluation and optimization of Linpack benchmark on ABCI","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"性能評価","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2018-12-10","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"株式会社富士通研究所"},{"subitem_text_value":"株式会社富士通研究所"},{"subitem_text_value":"株式会社富士通研究所"},{"subitem_text_value":"株式会社富士通研究所"},{"subitem_text_value":"株式会社富士通研究所"},{"subitem_text_value":"株式会社富士通研究所"},{"subitem_text_value":"株式会社富士通研究所"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Fujitsu Labratories Ltd.","subitem_text_language":"en"},{"subitem_text_value":"Fujitsu Labratories Ltd.","subitem_text_language":"en"},{"subitem_text_value":"Fujitsu Labratories Ltd.","subitem_text_language":"en"},{"subitem_text_value":"Fujitsu Labratories Ltd.","subitem_text_language":"en"},{"subitem_text_value":"Fujitsu Labratories Ltd.","subitem_text_language":"en"},{"subitem_text_value":"Fujitsu Labratories Ltd.","subitem_text_language":"en"},{"subitem_text_value":"Fujitsu Labratories Ltd.","subitem_text_language":"en"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_publisher":{"attribute_name":"出版者","attribute_value_mlt":[{"subitem_publisher":"情報処理学会","subitem_publisher_language":"ja"}]},"publish_status":"0","weko_shared_id":-1,"item_file_price":{"attribute_name":"Billing file","attribute_type":"file","attribute_value_mlt":[{"url":{"url":"https://ipsj.ixsq.nii.ac.jp/record/192841/files/IPSJ-HPC18167016.pdf","label":"IPSJ-HPC18167016.pdf"},"date":[{"dateType":"Available","dateValue":"2020-12-10"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-HPC18167016.pdf","filesize":[{"value":"437.3 kB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"14"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"9928bc5f-d885-4a36-afb3-b0978e3b3b38","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2018 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"本田, 巧"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"笠置, 明彦"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"福本, 尚人"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"大辻, 弘貴"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"土肥, 義康"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"田原, 司睦"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"中島, 耕太"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Takumi, Honda","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Akihiko, Kasagi","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Naoto, Fukumoto","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Hiroki, Ohtsuji","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Yoshiyasu, Doi","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Tsuguchika, Tabaru","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Kohta, Nakashima","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10463942","subitem_source_identifier_type":"NCID"}]},"item_4_textarea_12":{"attribute_name":"Notice","attribute_value_mlt":[{"subitem_textarea_value":"SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc."}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_18gh","resourcetype":"technical report"}]},"item_4_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"2188-8841","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"近年の HPC システムは，HPC アプリケーションに加え，高い演算性能が必要なディープラーニングへの需要も増加している．そのため，高い演算性能と電力効率を有する GPU を搭載したノードで構成される PC クラスタが増加している．GPU の搭載でノードの演算性能が大幅に向上した一方で，ノード間通信に用いられる InfiniBand や CPU-GPU 間のデータ転送に用いられる PCIe は相対的に性能向上が緩やかであるため，システム全体の性能バランスの観点では，通信性能がボトルネックになりやすい．そのため，GPU クラスタにおける大規模実行では通信処理の最適化が重要になる．本論文では，最新 GPU クラスタ向けに開発した HPL と通信の最適化，また，大規模 GPU クラスタにおける Linpack ベンチマークの特性について報告する．本研究の性能評価は，4352 基の NVIDIA Tesla V100 を有する GPU クラスタ “AI 橋渡しクラウド (ABCI) ” で行った．最適化を行った HPL を用いて性能評価を行ったところ，全ノードの実行で 19.88 PFLOPS の性能を達成し，ABCI は 2018 年 6 月の Top 500 において世界 5 位の実行性能を達成した．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"The demand for improving computing power increases in not only HPC field but also DL field. Thus, recent PC clusters are equipped with GPUs which have high computing power and energy efficiency. In contrast, communication performance such as InfiniBand and PCIe has been improved slowly, compared with computing power, and the data communication easily becomes a bottleneck in a case of execution of applications on a large GPU clusters. In this paper, we report optimization techniques which are applied for improving Linpack performance and the performance evaluation on ABCI (AI Briding Cloud Infrastracture). ABCI is a petascale GPU cluster which has 4352 NVIDIA Tesla V100 and 2176 Intel Xeon Gold 6148. Our HPL achieved 19.88 PFLOPS, and the score made ABCI the fifth fastest supercomputer on Top500 in June, 2018.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"8","bibliographic_titles":[{"bibliographic_title":"研究報告ハイパフォーマンスコンピューティング（HPC）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2018-12-10","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"16","bibliographicVolumeNumber":"2018-HPC-167"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"created":"2025-01-19T00:58:31.474183+00:00","updated":"2025-01-19T23:59:02.782032+00:00","id":192841}