{"created":"2026-02-25T05:53:56.285825+00:00","links":{},"updated":"2026-02-25T05:54:00.444997+00:00","id":2007963,"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:02007963","sets":["1164:2240:1771568311705:1771568419073"]},"path":["1771568419073"],"owner":"80578","recid":"2007963","title":["<i>N</i>体計算におけるGPUプログラミング手法比較：NVIDIA GH200/B200およびAMD MI300Aでの性能評価"],"pubdate":{"attribute_name":"PubDate","attribute_value":"2026-03-09"},"_buckets":{"deposit":"bded736a-75e1-40b1-b249-8ac3f4f74a86"},"_deposit":{"id":"2007963","pid":{"type":"depid","value":"2007963","revision_id":0},"owners":[80578],"status":"published","created_by":80578},"item_title":"<i>N</i>体計算におけるGPUプログラミング手法比較：NVIDIA GH200/B200およびAMD MI300Aでの性能評価","author_link":[],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"<i>N</i>体計算におけるGPUプログラミング手法比較：NVIDIA GH200/B200およびAMD MI300Aでの性能評価","subitem_title_language":"ja"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"性能評価とLLM活用技術","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2026-03-09","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"東京大学情報基盤センター"},{"subitem_text_value":"東京大学情報基盤センター"}]},"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/2007963/files/IPSJ-HPC26203047.pdf","label":"IPSJ-HPC26203047.pdf"},"date":[{"dateType":"Available","dateValue":"2028-03-09"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-HPC26203047.pdf","filesize":[{"value":"1.5 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":"14"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"7b7ade24-d5ea-4ae6-a4b1-cfa7af84cdb2","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2026 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"三木,洋平"}]},{"creatorNames":[{"creatorName":"塙,敏博"}]}]},"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":"日本国内においてGPUスパコンの導入事例が増加しており，また「富岳NEXT」への準備を進めるという観点においても科学技術アプリケーションのGPU対応が急務となっている．2024年までに導入されてきた国内のGPUスパコンは全てNVIDIA製GPUを搭載していたが，2025年にはAMD製GPUを搭載したシステムも導入されたことから，ベンダーニュートラルなGPUコード開発の必要性も高まっている．そこで本研究では，直接法N体計算コードをCUDA/HIP/SYCL/Kokkos/Solomonで実装・最適化し，NVIDIA GH200/B200およびAMD MI300Aという最新のGPU上で性能・移植性を比較した．NVIDIA GH200上ではSYCL実装をAdaptiveCppでコンパイルした時に最高性能となり，粒子数N=33 554 432における単位時間あたりの相互作用計算数は2.16×1012s-1であった．Kokkos実装についてもほぼ同等の性能を発揮し，CUDA C++実装を上回った．AMD MI300A上ではpacked FP32命令を用いた実装が有効であり，HIP実装では粒子数N=4 194 304における単位時間あたりの相互作用計算数が2.45×1012s-1に達した．SYCL実装をAdaptiveCppでコンパイルした時にもほぼ同等の性能が得られたことから，性能可搬プログラミングを用いた実装が，CUDA C++やHIP実装と同等以上の性能を発揮できると示された．また，NVIDIA B200上でのKokkos実装においては，粒子数N=33 554 432における性能が2.38×1012s-1という，NVIDIA GH200からの理論ピーク性能比にほぼ等しい性能向上が得られた．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"10","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":"2026-03-09","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"47","bibliographicVolumeNumber":"2026-HPC-203"}]},"relation_version_is_last":true,"weko_creator_id":"80578"}}