{"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00241667","sets":["1164:1579:11464:11813"]},"path":["11813"],"owner":"44499","recid":"241667","title":["SCALE-DGのModal Filteringにおけるループボディ分割の自動チューニング"],"pubdate":{"attribute_name":"公開日","attribute_value":"2024-12-09"},"_buckets":{"deposit":"78d2a19b-a519-4061-b1b9-6f3fcf46ff42"},"_deposit":{"id":"241667","pid":{"type":"depid","value":"241667","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"SCALE-DGのModal Filteringにおけるループボディ分割の自動チューニング","author_link":["665726","665732","665735","665737","665725","665731","665733","665738","665734","665724","665729","665739","665736","665727","665730","665728"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"SCALE-DGのModal Filteringにおけるループボディ分割の自動チューニング"},{"subitem_title":"Auto-tuning of Loop Body Splitting in Modal Filtering of SCALE-DG","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"自動最適化","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2024-12-09","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"名古屋大学大学院情報学研究科"},{"subitem_text_value":"理化学研究所計算科学研究センター"},{"subitem_text_value":"理化学研究所計算科学研究センター"},{"subitem_text_value":"理化学研究所計算科学研究センター"},{"subitem_text_value":"名古屋大学情報基盤センター"},{"subitem_text_value":"名古屋大学情報基盤センター"},{"subitem_text_value":"名古屋大学情報基盤センター"},{"subitem_text_value":"名古屋大学情報基盤センター"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Graduate School of Informatics, Nagoya University","subitem_text_language":"en"},{"subitem_text_value":"RIKEN Center for Computational Science","subitem_text_language":"en"},{"subitem_text_value":"RIKEN Center for Computational Science","subitem_text_language":"en"},{"subitem_text_value":"RIKEN Center for Computational Science","subitem_text_language":"en"},{"subitem_text_value":"Information Technology Center, Nagoya University","subitem_text_language":"en"},{"subitem_text_value":"Information Technology Center, Nagoya University","subitem_text_language":"en"},{"subitem_text_value":"Information Technology Center, Nagoya University","subitem_text_language":"en"},{"subitem_text_value":"Information Technology Center, Nagoya University","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/241667/files/IPSJ-ARC24259002.pdf","label":"IPSJ-ARC24259002.pdf"},"date":[{"dateType":"Available","dateValue":"2026-12-09"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-ARC24259002.pdf","filesize":[{"value":"2.2 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":"16"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"c101d14b-177f-4280-a7bf-8db67dedfaf3","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2024 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"任, 軒正博"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"河合, 佑太"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"富田, 浩文"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"西澤, 誠也"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"片桐, 孝洋"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"星野, 哲也"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"河合, 直聡"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"永井, 亨"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Xuanzhengbo, Ren","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Yuta, Kawai","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Hirofumi, Tomita","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Seiya, Nishizawa","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Takahiro, Katagiri","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Tetsuya, Hoshino","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Masatoshi, Kawai","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Toru, Nagai","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10096105","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":"2188-8574","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"将来の高解像度大気シミュレーションを念頭において，不連続ガラーキン法を用いた高精度大気力学コア (SCALE-DG) を開発している．SCALE-DG の数値安定化ための modal filtering の実装において，展開多項式の次数 (P) が大きい場合に，ループボディ内の積和演算を分割し演算器律速を削減することによって，さらなる高速化が可能である．しかし，最適な分割数は P に依存する．本研究では，SCALE-DG の modal filtering において，異なる多項式次数に対応する最適なループボディ分割数を特定するチューニングプロセスを提案した．スーパーコンピュータ「不老」の Type I サブシステムを用い，次数 P を 3 から 11 まで設定してチューニングプロセスを行った．結果として，P≤7 の場合にはループボディを分割することより性能が劣化することが確認された．一方，8≤P≤11 の場合にはループボディを 3 分割するのが最適であり，最大でそれぞれ1.15，1.20，1.25，および 1.26 倍の高速化が達成された．また，ループボディを分割する際には，各分割の長さを可能な限り均等に分配することが最適であることも確認された．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"SCALE-DG is a high-order fluid scheme developed for future high-resolution atmospheric simulations using the discontinuous Galerkin method (DGM). In the implementation of modal filtering for numerical stabilization in SCALE-DG, further acceleration is achievable by splitting the multiply-add operations in the loop body when the polynomial degree (P) is high, thereby reducing computational bottlenecks. However, the optimal number of splitting depends on the P. In this study, we proposed a tuning process for identifying the optimal number of loop body splitting for different polynomial degrees in modal filtering of SCALE-DG. The tuning process was executed from P = 3 to P = 11 on the supercomputer ”Flow” Type I subsystem. The results showed that the performance degraded with splitting the loop body when P ≤ 7. On the other hand, when 8 ≤ P ≤ 11, it was optimal to split the loop body into three parts, achieving speedups of up to 1.15, 1.20, 1.25, and 1.26 times, respectively. It was also confirmed that when dividing the loop body, it is optimal to distribute the length of each division as evenly as possible.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告システム・アーキテクチャ（ARC）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2024-12-09","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"2","bibliographicVolumeNumber":"2024-ARC-259"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":241667,"updated":"2025-01-19T07:35:10.441931+00:00","links":{},"created":"2025-01-19T01:46:24.382880+00:00"}