Development of a heterogeneous distributed-memory parallel iterative algorithm for practical use of ultra-large-scale finite element analysis

2016 Fiscal Year Final Research Report

• Project/Area Number: 24760062
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Engineering fundamentals
• Research Institution: Nagoya University
• Principal Investigator: Ogino Masao 名古屋大学, 情報基盤センター, 准教授 (00380593)
• Project Period (FY): 2012-04-01 – 2017-03-31
• Keywords: 領域分割法 / 並列有限要素法 / ヘテロジニアスコンピューティング / ハイパフォーマンスコンピューティング / 超大規模計算 / CPU-GPUハイブリッド並列処理

Outline of Final Research Achievements

For practical use of ultra-large-scale finite element analysis, a heterogenous distributed-memory parallel iterative algorithm based on the domain decomposition method has been studied. In particular, an optimization implementation of the DDM on the GPGPU and a many-core processor, a Scaled-BDD method for multiple-material model, and a MINRES-like_CS method for solving complex symmetric linear systems arising from electromagnetic analysis have been developed. As the result, a high-frequency electromagnetic field analysis with tens of billions degrees of freedom model has been successfully solved within a practical time limit.

Free Research Field

計算力学