Author(s): Shanghong Zhang, Wenda Li, Zhu Jing
Linked Author(s): shanghong zhang
Keywords: Dam-break flow, parallel computation, graphics processing unit, OpenACC, OpenMP
Abstract: High performance calculations are of great importance to the simulation of dam-break events, as accuracy and accelerated speed are the key factors in the process of dam-break flow modeling. Based on the finite volume method, this paper established a high performance dam-break flow simulation model. An explicit scheme was used to discretize the 2D shallow water control equations and Roe's approximate Riemann solution of the finite volume method was adopted for the interface flux of the grid cells. A graphics processing unit (GPU) -based parallel method, OpenACC, and Open Multi-Processing (OpenMP) parallel mode are used to realize parallel computing, respectively. Because an explicit discrete technique is used to solve the governing equations, and there is no correlation between the grid calculations in a single time step, the parallel dam-break model can be easily realized by adding OpenACC and OpenMP directives to the loop structure of the grid calculations. To analyze the performance of the model, we considered the Pangtoupao flood storage area in China using a multi-core computer with an Nvidia Tesla K20c card and two different grid division schemes. By carefully studying the implementation method and optimization of data transportation in the parallel algorithm, a speedup factor of 17. 76 can be achieved with OpenACC and a speedup factor of 8. 6 was achieved with the conventional OpenMP parallel mode on a 16-kernel computer. The results demonstrate that the optimized feature settings greatly influence the degree of speedup and models involving larger numbers of calculations exhibit greater efficiency and higher speedup factors. In addition, both of the OpenACC and OpenMP parallel modes are found to have good portability, making it easy to implement parallel computation from the original serial model. And, it is possible to simulate dam-break flows in large-scale watersheds on a single computer with parallel computing
Year: 2017