Gpu Parallel Computing Applications And Optimization Of Particle Sedimentation Lattice Boltzmann Simulation

Particle sedimentation is a very common phenomenon in nature, widespread in many areas, such as industrial applications, life sciences, environmental science and medical science. Therefore, in recent years, the study of phenomenon of particle sedimentation attracted the attention of many scholars. As the particle sedimentation involves complex calculations, people can not use theoretical approach to solving this problem,and the experimental method has also been all sorts of obstacles. With the rise of numerical methods, to simulate the phenomenon of particle sedimentation brings hope. Currently, numerical simulation approach applied in the field of fluid dynamics is lattice Boltzmann method (LBM).LBM is a new computational method of fluid dynamics developed in recent years. Its algorithm is simple, easy to handle complex boundary, and also has a natural parallelism, very suitable for massively parallel computing. Therefore, it is widely used for a variety of complex fluid dynamics simulation issues. It is widely used for numerical simulation of complex fluid dynamics problems. About LBM, in the second chapter of this paper are briefly introduced. In this paper, the two-dimensional of single-particle dynamics model was established with LBM method, which was used to simulate two-dimensional single particle sedimentation in the fluid.Through the use of LBM, the study of particle sedimentation phenomenon has been much progress, but most just to achieve the simulation of various particle sedimentation. But no one consider the efficiency of its simulation. Due to the complexity of the model of particle sedimentation, large amounts of data, When model has large-scale, the simulation will be spent a long time and leading to inefficiencies. In this rapid develop society, efficiency is a very important factor, and the phenomenon of particle sedimentation can be widely used, therefore It is necessary to accelerate the simulation of model to improve its efficiency. In this paper, the research focus is how to improve the efficiency of the simulation of particle sedimentation.Although LBM has a natural parallelism, but only some improements have made with respect to the traditional numerical simulation methods. To make the simulation more efficient, we use the GPU parallel computing methods based on CUDA, develop and became popular rapidly in recent year. The fluid simulation was implement successfully by using the "CPU+GPU" heterogeneous model and CUDA programming, on the basis of LBM model. It means to achieve the best performance with the combination of the parallelism of LBM and GPU. GPU and CUDA are briefly described in Chapter III. This paper proves the feasibility and efficiency of parallel computing methods based on GPU with two basic examples--Poisueille flow and Cavity flow. Then, the parallel algorithm of numerical simulation of particle sedimentation movement based on LBM in GPU was designed and implemented under CUDA framework. On the ordinary personal computer, The CPU and GPU were used to simulate seperately. Experimental results show that the parallel computing of numerical simutation based on LBM particle sedimentation in GPU is feasible entirely. The simulation results fully consistent with the results of the simulation on the CPU, and received a very impressive acceleration.Although the GPU-based parallel computing greatly improves the efficiency of the program, but we are not satisfied with this, on this basis can be optimized through a variety of methods. In this paper, we use several important optimization methods of CUDA programming that described in Chapter IV, such as dimensions divided optimization, memory access optimization, instruction stream optimization and comprehensive optimization methods, for the further optimization of the particle sedimentation GPU-based simulation program. Experimental results show that the efficiency of the optimized program has been further improved, and the comprehensive optimization has most obvious effect and won the72-fold speedup. The achievement of accelerated the LBM simulation model of particle sedimentation, laid the foundation for a more widely used in real life, and also contributed to the GPU general purpose parallel computing technology in the development process used in more practical problems.