Par Allel Optimization Of Seismic Wave Simulation By Finite Differ Ence Method

Seismic wave simulation is an indispensable part of Earth Science research.In geophysics,numerical simulation reveals some physical field information,which is not easy to obtain when measuring,so the behavior of the field and underground structure can be better understood by numerical simulation.For smaller models,a single CPU can be used to simulate seismic wave field.However,when the scale is large,the operation efficiency is far from meeting the requirements of industrial application.Moreover,when the model is huge,the memory of a single CPU is often difficult to meet its computing storage requirements.Parallel computing is an effective way to solve such huge computational load or large memory requirement of single CPU.Therefore,it is of great theoretical significance and application value to study the parallel optimization of finite difference seismic wave simulation.The parallel algorithm of finite difference method in seismic simulation is studied in this paper.The solution mainly includes the following parts:(1)By analyzing the data dependence and parallelism of the finite difference method,the parallel method of domain decomposition is used to parallelize the finite difference method for seismic wave simulation,so as to solve the memory deficiency of the large-scale finite difference method for seismic wave simulation.(2)By testing different regions of multi-node,according to the impact of the scale of the problem and hardware resources on the running time of the program,it is found that the running time of the program varies in different regions.Therefore,it is necessary to choose a suitable partition method to reduce the total running time.The experimental results show that when the model area of each process is the same size,the running time of single node and multi-node is roughly the same.Therefore,a method of predicting multi-node partition based on single node test is proposed to optimize the parallel implementation of finite difference seismic wave simulation.(3)By analyzing the iterative computation method of the traditional parallel finite difference method,aiming at the shortcomings of the traditional parallel finite difference method that its computation and trust can only be executed sequentially and independently,the non-blocking communication method is chosen to hide the communication time.At the same time,the computational characteristics of the finite difference method are analyzed,and the method of judging the boundary points of non-blocking communication implemented in this paper is designed.Finally,thenon-blocking communication Iterative computation method of blocking communication is presented,and the optimization effect of this method is verified by experimental comparison.(4)The algorithm implemented in this paper is tested on an over-computing platform.Through the analysis of the experimental results,it is concluded that the parallel implementation method of finite difference seismic wave simulation in this paper is closer to the parallel linear acceleration ratio than the traditional method,and has better scalability.At the same time,the running time of the program is reduced to a certain extent.