2026-03-10T03:34:56Z https://ipsj.ixsq.nii.ac.jp/oai

oai:ipsj.ixsq.nii.ac.jp:00218963 2025-01-19T14:58:20Z 1164:2240:10902:10971

A Parallel-in-Time Method for Compressible Fluid Explicit Simulation A Parallel-in-Time Method for Compressible Fluid Explicit Simulation Yen-Chen, Chen Kengo, Nakajima Yen-Chen, Chen Kengo, Nakajima CFD High-Performance Computing (HPC) methods for time-dependent problems reach an acceleration in space as the problem size grows. This restriction leads to the development of parallel-in-time (PinT) methods. Despite many PinT methods have been introduced, very few PinT methods have been tested with explicit schemes to our best knowledge. This research introduces a PinT method that works with explicit schemes. This research constructs a multi-layer hierarchy in time and space and solves it through coarse to fine layers. The proposed Cascadic Parareal method is optimized based on the number of available cores to improve the efficiency of parallel-in-time solvers with a limited number of processors. The numerical experiment solves for a compressible fluid simulation around a cylinder. The research result shows that the proposed parallel-in-space/time (PinST) method could solve faster than traditional spatial parallelization with less than 100 cores. High-Performance Computing (HPC) methods for time-dependent problems reach an acceleration in space as the problem size grows. This restriction leads to the development of parallel-in-time (PinT) methods. Despite many PinT methods have been introduced, very few PinT methods have been tested with explicit schemes to our best knowledge. This research introduces a PinT method that works with explicit schemes. This research constructs a multi-layer hierarchy in time and space and solves it through coarse to fine layers. The proposed Cascadic Parareal method is optimized based on the number of available cores to improve the efficiency of parallel-in-time solvers with a limited number of processors. The numerical experiment solves for a compressible fluid simulation around a cylinder. The research result shows that the proposed parallel-in-space/time (PinST) method could solve faster than traditional spatial parallelization with less than 100 cores. technical report 情報処理学会 2022-07-20 application/pdf 研究報告ハイパフォーマンスコンピューティング（HPC） 25 2022-HPC-185 1 7 2188-8841 AN10463942 https://ipsj.ixsq.nii.ac.jp/record/218963/files/IPSJ-HPC22185025.pdf eng