Research On Application Of GPU Parallel Computing And CUDA Programming In Environmental Engineering

With the rapid development of social economy,the contradiction between water supply and demand has become increasingly prominent,resulting in many environmental problems.Groundwater numerical simulation technology is very important for studying the geological problems in the environment,as it can not only predict the groundwater level and rationally develop and plan the utilization of groundwater resources,but solves various geological and environmental problems.Groundwater numerical simulation describes the change law of groundwater level with time and space.As the groundwater numerical simulation is well researched,the research model has finer grid granularity,the long time span and the increasingly expanding research scale,which results in production of massive data.Traditional serial simulation techniques are increasingly unable to meet this high-precision grid partitioning and simulation of large-scale groundwater flow model.Parallel technology can effectively deal with massive data generated in the simulation process.It can effectively reduce the simulation time to improve the simulation efficiency,which provides a strong guarantee for the researches on large-scale groundwater flow simulation.As the generalized minimal residual method GMRES(m)is not limited to the properties of the matrix and has a strong adaptability to ordinary matrixes,it can be widely applied to the solution of large linear equations.The high quality preconditioners are selected and combined with GMRES(m)algorithm,which can effectively improve the spectral norm of the coefficient matrix and accelerate the convergence speed of the algorithm.In this paper,the GPU parallel technology platform is used to parallelize the GMRES(m)and applied to iterative solution in the groundwater simulation system,so as to achieve effective processing of massive data by the groundwater numerical simulation system.Based on application of GPU parallel technology in the groundwater numerical simulation system,the following works are finally completed:(1)The parallelism of the GMRES(m)algorithm is analyzed.The CUDA-basedcuBLAS and cuSparse function libraries are used to parallelize the GMRES(m)algorithm,which is then applied to the iterative solution of the groundwater numerical simulation system to speed up the solution of the linear equations.(2)The jacobi and ILU0 preconditioners are chosen and respectively combined with the parallel GMRES(m)algorithm to accelerate the convergence speed of iterative solution of the parallel GMRES(m)algorithm in the groundwater numerical simulation system.(3)Different scales of the groundwater numerical models are selected to test GPU-based parallel platform and parallelize the GMRES(m)algorithm to speed up iteration solution in the groundwater simulation system.The parallelization efficiency and reliability of the GMRES(m)algorithm is verified.