Runtime Optimization Technologies for Communications of CA

• Project/Area Number: 21700036
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Single-year Grants
• Research Field: Software
• Research Institution: Kyushu University
• Principal Investigator: NANRI Takeshi 九州大学, 情報基盤研究開発センター, 准教授 (70284578)
• Project Period (FY): 2009 – 2011
• Project Status: Completed (Fiscal Year 2011)
• Budget Amount: ¥3,640,000 (Direct Cost: ¥2,800,000、Indirect Cost: ¥840,000); Fiscal Year 2011: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000); Fiscal Year 2010: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000); Fiscal Year 2009: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
• Keywords: 並列処理 / 動的最適化 / 並列プログラミング言語 / 通信最適化 / 動的最適化技術 / HPC / CAF

Research Abstract

Parallel programs on parallel computers invoke communications among processes to progress in their executions. In this project, a technique that choose the appropriate algorithm of such communications according to the status of the program at runtime is introduced, so that it can reduce the time for communication and enable efficient execution of the programs. Results of experiments showed the effectiveness of the proposed technique.

Research Products

• [Presentation] Efficient Runtime Algorithm Selection of Collective Communicationwith Topology-Based PerformanceModels (2012)
  • Author(s): Takeshi Nanri and Motoyoshi Kurokawa
  • Organizer: The 2012 International Conference on Parallel and Distributed Processing Techniques and Applications
  • Place of Presentation: To appear
• [Presentation] ランク配置に応じた集団通信アルゴリズム動的選択技術の提案 (2012)
  • Author(s): 南里豪志、黒川原佳
  • Organizer: 第133回ハイパフォーマンスコンピューティング研究会
• [Presentation] ランク配置に応じた集団通信アルゴリズム動的選択技術の提案 (2012)
  • Author(s): 南里豪志
  • Organizer: 第133回ハイパフォーマンスコンピューティング研究会
  • Place of Presentation: 有馬ビューホテルうらら(兵庫)
• [Presentation] 通信ライブラリにおける実行時自動チューニング技術 (2011)
  • Author(s): 南里豪志
  • Organizer: 第3回自動チューニング技術の現状と応用に関するシンポジウム
  • Place of Presentation: 東京大学小柴ホール(東京都)(招待講演)
  • Year and Date: 2011-12-05
• [Presentation] Task Allocation Method for Avoiding Contentions by the Information of Concurrent Communications (2011)
  • Author(s): Yoshiyuki Morie
  • Organizer: IASTED Conference on Parallel and Distribuited Computing and Networks 2011
  • Place of Presentation: Innsbruck, Austria
  • Year and Date: 2011-02-17
• [Presentation] 充填率を考慮したジョブスケジューラによるランク配置に対する集団通信アルゴリズム自動選択技術の効果 (2011)
  • Author(s): 南里豪志
  • Organizer: 第14回環瀬戸内応用数理研究部会シンポジウム
  • Place of Presentation: 岡山市
  • Year and Date: 2011-01-23
• [Presentation] Effect of Dynamic Algorithm Selection of All-to-All Communication on Environments with Unstable Network Speed (2011)
  • Author(s): Takeshi Nanri and Motoyoshi Kurokawa
  • Organizer: International Conference on High Performance Computing & Simulation
• [Presentation] A Method for Predicting a Penalty of Contentions by Considering Priorities of Routing among Packets on Direct Interconnection Network (2011)
  • Author(s): Yoshiyuki Morie, Takeshi Nanri, Ryutaro Susukita and Koji Inoue
  • Organizer: in Proceedings of the International Joing Conference on Computational Sciences and Optimization
• [Presentation] ask Allocation Method for Avoiding Contentions by the Information of Concurrent Communications (2011)
  • Author(s): Yoshiyuki Morie, Takeshi Nanri and Motoyoshi Kurokawa
  • Organizer: Proceedings of IASTED PDCN
• [Presentation] Effect of Dynamic Algorithm Selection of All-to-All Communication on Environments with Unstable Network Speed (2011)
  • Author(s): Takeshi Nanri
  • Organizer: 2011 International Conference on High Performance Computing and Simulation
  • Place of Presentation: Istanbul, Turkey
• [Presentation] 通信と計算の負荷を考慮した並列疎行列ベクトル積の動的負荷分散技術 (2010)
  • Author(s): 南里豪志
  • Organizer: 情報処理学会SWoPP2010
  • Place of Presentation: 金沢市文化ホール
  • Year and Date: 2010-08-04
• [Presentation] Runtime Load-balancing Technique for Sparse MatrixVector Multiplication (2010)
  • Author(s): Kenichiro Kusaba, Takeshi Nanri and Seiji Fujino
  • Organizer: nternational Workshop on Innovative Architecture
• [Presentation] Runtime Load-Balancing Technique for Sparse Matrix-Vector Multipulication (2010)
  • Author(s): Takeshi Nanri
  • Organizer: International Workshop on Innovative Architecture
  • Place of Presentation: Kona, USA
• [Presentation] ARobustDynamicOptimizationfor MPI AIItoall Operation (2009)
  • Author(s): yacintheNzigouMamadou, TakeshiNanri, andKazuakiMurakami
  • Organizer: 8th Internationa lHeterogeneity in Computing Workshop
• [Presentation] A Dynamic Solution for Efficient MPI Collective Communications (2009)
  • Author(s): yacinthe Nzigou Mamadou, FengLong Gu, Vivien Oddou, Takeshi Nanri, Kazuaki Murakami
  • Organizer: Internationalworkshop on HPC and Grid Applications
• [Presentation] A Dynamic Solution for Efficent MPI collective Communications (2009)
  • Author(s): Hyacinthe Nzigou Mamadou
  • Organizer: International Workshop on HPC and Grid Applications
  • Place of Presentation: Sanya, China
• [Presentation] 階層型並列計算機向けPAGMEつきCG法の実装と性能解析 (2009)
  • Author(s): 南里豪志
  • Organizer: 計算工学講演会
  • Place of Presentation: 東京
• [Presentation] A Robust Dynamic Optimization for MPI Alltoall Operation (2009)
  • Author(s): Takeshi Nanri
  • Organizer: International Heterogeneity in Computing Workshop
  • Place of Presentation: Rome, Italy