{"created":"2025-01-19T00:56:35.868800+00:00","updated":"2025-01-20T01:07:39.569259+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00190641","sets":["1164:1867:9409:9525"]},"path":["9525"],"owner":"11","recid":"190641","title":["GPUミドルウェア”Victream”のシミュレーションによる性能評価"],"pubdate":{"attribute_name":"公開日","attribute_value":"2018-07-23"},"_buckets":{"deposit":"430d22b9-3d1a-489d-9aa8-4ba5cb4b0637"},"_deposit":{"id":"190641","pid":{"type":"depid","value":"190641","revision_id":0},"owners":[11],"status":"published","created_by":11},"item_title":"GPUミドルウェア”Victream”のシミュレーションによる性能評価","author_link":["436847","436845","436841","436844","436840","436843","436846","436842"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"GPUミドルウェア”Victream”のシミュレーションによる性能評価"},{"subitem_title":"Simulation Evaluation of Performance of GPU Middleware ”Victream”","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"GPU","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2018-07-23","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"NECデータサイエンス研究所／東京大学生産技術研究所"},{"subitem_text_value":"NECデータサイエンス研究所"},{"subitem_text_value":"NECデータサイエンス研究所"},{"subitem_text_value":"東京大学生産技術研究所"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"NEC / Institute of Industrial Science, the University of Tokyo","subitem_text_language":"en"},{"subitem_text_value":"NEC","subitem_text_language":"en"},{"subitem_text_value":"NEC","subitem_text_language":"en"},{"subitem_text_value":"Institute of Industrial Science, the University of Tokyo","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/190641/files/IPSJ-OS18144016.pdf","label":"IPSJ-OS18144016.pdf"},"date":[{"dateType":"Available","dateValue":"2020-07-23"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-OS18144016.pdf","filesize":[{"value":"11.9 MB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"11"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"1d04bec8-3e31-4253-b925-551f095b2838","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":"鈴木, 順"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"林, 佑樹"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"荒木, 拓也"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"喜連川, 優"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Jun, Suzuki","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Yuki, Hayashi","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Takuya, Araki","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Masaru, Kitsuregawa","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10444176","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":"2188-8795","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"Victream は著者らが [1] で提案した複数の GPU を用いた Out-of-Core 処理向けミドルウェアである．Out-of-Core 処理では GPU (Graphics Processing Unit) のメモリ容量より大きなデータを分割し，GPU とホストのメインメモリの間で入れ替えながら処理を行う．そのため GPU のデータ I / O が性能ボトルネックとなる．このボトルネックの緩和のため，Victream は GPU へのデータのプリフェッチと，GPUへのデータ I / O 量最小化を同時に実現する．そのため Victream スケジューラは動的でヒューリスティックなスケジューリング手法を採用している．しかし，ヒューリスティック手法は必ずしも最良のスケジュールを保証しない．そこで本稿では，与えられた計算に対して可能なスケジュールを Brute-Force シミュレーションで全探索し，得られた最良のスケジュールと Victream のスケジュールの計算時間を比較することで Victream の性能を評価する．結果，Victream は最良のスケジュールが実現する最短の計算時間ではないが，最短の計算時間に対する増加を 10% 以内に抑えた計算性能が実現できることがわかった．本稿ではさらに，Victream をホストと GPU を分離して計算プラットフォームの高い運用効率を実現する I / O デバイス分離のアーキテクチャに適用し，Victream の有効性を検証した．I / O デバイス分離のアーキテクチャでも性能ボトルネックはホストと GPU の間のデータ I / O となる．評価の結果，Victream は I / O デバイス分離のアーキテクチャで従来手法より効率的に GPU の演算リソースを使用でき，用いる GPU 数を増加させた場合も高い計算性能の向上が得られることがわかった．評価では，Victream は従来手法に対し最大 296% の性能向上を実現した．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"9","bibliographic_titles":[{"bibliographic_title":"研究報告システムソフトウェアとオペレーティング・システム（OS）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2018-07-23","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"16","bibliographicVolumeNumber":"2018-OS-144"}]},"relation_version_is_last":true,"weko_creator_id":"11"},"id":190641,"links":{}}