| Item type |
SIG Technical Reports(1) |
| 公開日 |
2019-07-17 |
| タイトル |
|
|
タイトル |
Parallel-in-Space/Time Method for Explicit Time-Marching Scheme |
| タイトル |
|
|
言語 |
en |
|
タイトル |
Parallel-in-Space/Time Method for Explicit Time-Marching Scheme |
| 言語 |
|
|
言語 |
eng |
| キーワード |
|
|
主題Scheme |
Other |
|
主題 |
数値計算 |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_18gh |
|
資源タイプ |
technical report |
| 著者所属 |
|
|
|
The University of Tokyo |
| 著者所属 |
|
|
|
The University of Tokyo |
| 著者所属(英) |
|
|
|
en |
|
|
The University of Tokyo |
| 著者所属(英) |
|
|
|
en |
|
|
The University of Tokyo |
| 著者名 |
Yen-Chen, Chen
Kengo, Nakajima
|
| 著者名(英) |
Yen-Chen, Chen
Kengo, Nakajima
|
| 論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
Numerical PDE solvers require heavy computation power to solve and efficient parallel processing to accelerate. The traditional methods parallelize only in the spatial dimensions but not the time dimension. However, this limited the scalability we can do by parallel computation. Therefore, several Parallel-in-Space/Time (PinST) methods are proposed to accelerate PDE solvers further. One of the famous PinST methods is MGRIT method[1]. The MGRIT method takes advantages of the multigrid method and applied it on both the space dimensions and the time dimension, which has been proved to work well on linear problems. Another method is the Time Segment Correction (TSC), which works more stable on highly non-linear problems. However, the number of converge iteration for TSC grows with time dimension size, therefore requires more computation on large-scale problems. In the present work, we propose a new PinST method targetting explicit time-marching schemes. |
| 論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
Numerical PDE solvers require heavy computation power to solve and efficient parallel processing to accelerate. The traditional methods parallelize only in the spatial dimensions but not the time dimension. However, this limited the scalability we can do by parallel computation. Therefore, several Parallel-in-Space/Time (PinST) methods are proposed to accelerate PDE solvers further. One of the famous PinST methods is MGRIT method[1]. The MGRIT method takes advantages of the multigrid method and applied it on both the space dimensions and the time dimension, which has been proved to work well on linear problems. Another method is the Time Segment Correction (TSC), which works more stable on highly non-linear problems. However, the number of converge iteration for TSC grows with time dimension size, therefore requires more computation on large-scale problems. In the present work, we propose a new PinST method targetting explicit time-marching schemes. |
| 書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AN10463942 |
| 書誌情報 |
研究報告ハイパフォーマンスコンピューティング（HPC）
巻 2019-HPC-170,
号 37,
p. 1-5,
発行日 2019-07-17
|
| ISSN |
|
|
収録物識別子タイプ |
ISSN |
|
収録物識別子 |
2188-8841 |
| Notice |
|
|
|
SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc. |
| 出版者 |
|
|
言語 |
ja |
|
出版者 |
情報処理学会 |
IPSJ-HPC19170037.pdf (1.5 MB)
