Using Generalized Finite Difference Method and GPU (Graphic Processing Unit) Parallel Computation to Develop a Highly Efficient Model for Shallow Water Equations

View Statistics Email Alert RSS Feed

Information

Details

Project title

Code/計畫編號

MOST104-2221-E019-036

Translated Name/計畫中文名

結合廣義有限差分法與圖形處理器平行計算以發展高效率之淺水波方程式模擬模式

Project Coordinator/計畫主持人

Chia-Ming Fan

Funding Organization/主管機關

National Science and Technology Council

Department/Unit

Department of Harbor and River Engineering

Website

https://www.grb.gov.tw/search/planDetail?id=11580414

Year

2015

Start date/計畫起

01-08-2015

Expected Completion/計畫迄

31-07-2016

Bugetid/研究經費

669千元

ResearchField/研究領域

土木水利工程

"本計畫為延續前一年度計畫執行成果，將前一年度研發完成之淺水波方程式電腦模擬模式，進一步結合圖形處理器平行計算技術，以降低電腦硬體投資金額為前提，開發一套淺水波方程式之平行化電腦模擬模式，以便將來提供給工程規劃設計與學術研究使用。近年來由於氣候變遷等因素，在世界各地發生很多自然災害，此類問題都與水的流動有密切關係，因此我們迫切需要發展一套準確且快速的淺水波方程式模擬模式，以幫助我們了解不同設定條件下的流體流動行為，所以本計畫預計開發一套高效率之淺水波方程式平行計算模擬模式。在前一年度的計畫中，我們已經採用廣義有限差分法與顯式積分法發展一套淺水波方程式之無網格法模擬模式，此一模式可以避免網格產生與數值積分的耗時工作，也可以快速且準確的模擬淺水波方程式之流體流動現象。然而，有鑑於近年來電腦圖形處理器的快速發展，在單一圖形處理器上就會有約數百顆計算核心，若能夠有效率的使用這些計算核心，必能發揮非常高的電腦模擬效能。因次本計畫預計以圖形處理器為主，將前一年度開發完成之淺水波方程式模式平行化，並探討本模式之平行計算效能，進行綜合討論分析，以提供給未來工程開發與學術研究使用。""In this project, we plan to develop a parallel computation model for shallow water equations (SWE) by combining the meshless numerical model for SWE, which is completed in the previous project, and the parallel computation of graphic processing unit (GPU). By the consideration of limited budget for computer hardware, a parallel computation model will be developed in this project and could be provided to engineering design and academic research in the future. Due to the climate change and other effects, we have faced several disasters from nature in the recent years. In order to realize these fluid-related events, we urgently need an efficient and accurate simulation model for SWE to understand the flow fields under different conditions. So, in this project we plan to develop a highly efficient parallel computation model for analyzing SWE. During the previous project, a meshless model for SWE has been developed by combining the generalized finite difference method and the explicit Runge-Kutta scheme. Since the time-consuming tasks of mesh generation and numerical quadrature are avoided, the developed model is very efficient and useful. However, in consideration of quick developments of GPU in the past decade, the simulation efficiency of SWE model can be greatly improved if hundreds of cores in a single GPU can be perfectly adopted. Therefore, in this project, we will combine the GPU parallel computation and the meshless model for SWE to form a highly efficient simulation model. In addition, we will systematically examine the parallel efficiency of the developed model and compared the results with others. The parallel model for SWE, developed in this project, can be provided to engineering design and academic research in the future."

Keyword(s)

圖形處理器
平行計算
淺水波方程式
無網格法
高效率
GPU (graphic processing unit)
parallel computation
shallow water equations
meshless method
highly efficient model

DSpace CRIS

Using Generalized Finite Difference Method and GPU (Graphic Processing Unit) Parallel Computation to Develop a Highly Efficient Model for Shallow Water Equations

Details

Description