Strategies and Performance Evaluation of Parallel Computation in Solving the Laplace Equation

• Item type: JInfP(1)
• 公開日: 1983-12-20
• タイトル: Strategies and Performance Evaluation of Parallel Computation in Solving the Laplace Equation
• タイトル / 言語: en
• タイトル: Strategies and Performance Evaluation of Parallel Computation in Solving the Laplace Equation
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_6501
• 資源タイプ: journal article
• 著者所属: Department of Information Science Faculty of Engineering Tokushima University
• 著者所属: Department of Industrial Control Systems Omika Works Hitachi Co. Ltd.
• 著者所属: Heavy Apparatus Engineering Laboratory Toshiba Inc.
• 著者所属(英): en
• 著者所属(英): Department of Information Science, Faculty of Engineering, Tokushima University
• 著者所属(英): en
• 著者所属(英): Department of Industrial Control Systems, Omika Works, Hitachi Co., Ltd.
• 著者所属(英): en
• 著者所属(英): Heavy Apparatus Engineering Laboratory, Toshiba Inc.
• 著者名: Yoshizo, Takahashi; Yoshihiro, Nobutomo; Toshikazu, Kawamura
• 著者名(英): Yoshizo, Takahashi; Yoshihiro, Nobutomo; Toshikazu, Kawamura
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: The factors which have influence on the efficiency of parallel computation are the scheme of problem decomposition the processor interconnection and the OS overhead. In this paper the scheme of problem decomposition of parallel computation in solving the Laplace equation numerically is studied from two different viewpoints the strategy of partitioning mesh-points into blocks and the strategy for assignment of these blocks to the processors. Two matrices called the partition matrix and the distance matrix are introduced and an algorithm to give an optimum processor assignment to the blocks in the sense that it results in minimum data transmission in the whole system is presented. The performance of parallel computation in solving a Lapalce equation is evaluated by developing a software simulator of CORAL system which is a binary tree processor network. The result in comparison with serial computation shows that a speed-up ratio of 100 is attained with 300 processors in a binary tree.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: The factors which have influence on the efficiency of parallel computation are the scheme of problem decomposition, the processor interconnection, and the OS overhead. In this paper, the scheme of problem decomposition of parallel computation in solving the Laplace equation numerically is studied from two different viewpoints, the strategy of partitioning mesh-points into blocks and the strategy for assignment of these blocks to the processors. Two matrices called the partition matrix and the distance matrix are introduced and an algorithm to give an optimum processor assignment to the blocks in the sense that it results in minimum data transmission in the whole system is presented. The performance of parallel computation in solving a Lapalce equation is evaluated by developing a software simulator of CORAL system which is a binary tree processor network. The result, in comparison with serial computation, shows that a speed-up ratio of 100 is attained with 300 processors in a binary tree.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AA00700121
• 書誌情報: Journal of Information Processing; 巻 5, 号 4, p. 239-246, 発行日 1983-12-20
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 1882-6652
• 出版者 / 言語: ja
• 出版者: 情報処理学会