Study on acceleration method of the flood model based on GPU
KANG Yongde
KANG Aize
HOU Jingming
XU Erwen
REN Xiaofeng
Abstract：Due to the dual effects of global climate warming and extreme rainstorms,flood disasters occur frequently.It is very important to improve the computational efficiency of flood model for real-time simulation and prediction of flood.However,the huge amount of calculation brought by refined flood simulation makes it difficult to realize real-time simulation calculation,which cannot meet the needs of real-time calculation of simulation results and release of flood warning.An efficient and high-precision full hydrodynamic numerical model based on GPU acceleration technology is constructed,and the computational efficiency acceleration ratio of GPU and CPU in flood simulation is quantitatively studied.The results show that:Under the same settings,NVIDIA Tesla P100-PCIE has the best computational efficiency for other types of computing engines.With the same DEM grid resolution and different rainfall return periods,the GPU computing efficiency increases with the increase of rainfall return period,and the GPU/CPU parallel computing efficiency acceleration ratio is 1.25～16.28 times.When the recurrence period of rainfall is the same,the higher the resolution accuracy of the DEM grid is,the more significant the GPU acceleration efficiency is.When the grid resolution is 3m and 5m,the computational efficiency of NVIDIA GeForce GTX 980Ti is 4.32 and 3.26 times higher than that of CPU(single core),while NVIDIA Tesla P100-PCIE can increase by 16.28 and 7.86 times respectively.In summary,while ensuring better simulation accuracy,the finer the DEM grid resolution,the higher the GPU acceleration calculation efficiency.
Keywords：flood disastertwo-dimensional hydrodynamic modelGPUcomputational efficiency
Publication Date：2025-06-30
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：9( 64-72 )