Development of stencil calculation and communication model to achieve high scalability on massively parallel computation
| Project/Area Number |
18K11336
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 60100:Computational science-related
|
|---|
| Research Institution | Kyoto University |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
南里 豪志 九州大学, 情報基盤研究開発センター, 准教授 (70284578)
|
|---|
| Project Period (FY) |
2018-04-01 – 2022-03-31
|
| Project Status |
Completed (Fiscal Year 2021)
|
| Budget Amount *help |
¥4,290,000 (Direct Cost: ¥3,300,000、Indirect Cost: ¥990,000)
Fiscal Year 2020: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2019: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
Fiscal Year 2018: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
|
|---|
| Keywords | 並列計算 / 高性能計算 / MHDシミュレーション / スケーラビリティ / ステンシル計算 / 超並列計算 / エクサスケール |
|---|
| Outline of Final Research Achievements |
The purpose of this study is to develop a stencil calculation and communication model without scalability attenuation in exascale computing environment, and to develop the Halo communication function used there. First, in the stencil simulation, we developed a model that divides threads into "calculation" and "calculation and communication that require communication (including pack / unpack of communication data)". As a result, there is no need for synchronization to know when the communication has ended, and deterioration of parallel performance can be avoided. Then, we summarized a Halo function group from the model so that it could be easily used in other applications. These performances were measured in an environment using 2000 nodes, and high scalability was confirmed.
|
| Academic Significance and Societal Importance of the Research Achievements |
近年の計算機はプロセッサ単体の性能を周波数向上が難しいことから上がられず、目に-コア化や並列ノード数増加により性能向上を達成している。この様な中で並列計算時にスケーラビリティが低下すると、並列数増加分の性能向上が得られない。本研究の成果により高並列時にもスケーラビリティが減少しない手法が開発され、近年の計算機をより効率良く利用することが可能となる。
また、ステンシル計算はある空間を扱う計算で用いられる計算であり、スパコンを利用するような大規模計算アプリケーションの大部分を占めるため、幅広い領域へ貢献することが推測され、学術的意義が深い。
|
Report
(5 results)
Research Products
(18 results)
