OpenMP-accelerated element-free reverse-time migration

Zhen ZHOU,Xiao-feng JIA

DOI: https://doi.org/10.3969/j.issn.1001-1749.2016.06.17

2016-01-01

Abstract:The wave-equation-based method was widely used in seismic modeling and imaging in the past years.Many numerical strategies such as the finite element method (FEM)and the finite difference method (FDM)are developed in solving seismic wave equations.However,both methods have some shortcomings of either accuracy or computational cost in practice. In fact,element-free method (EFM)has been applied in seismic modeling and seismic migration imaging.Compared with FEM and FDM,EFM has unique advantages:it doesn't need to be divided a large number of grid in advance and it satisfies local fitting because only the information of the nodes and the boundary of the study area are required in computation.However, there are many problems that EFM is applied in seismic modeling and seismic migration imaging in present,e.g.the problems of storage and computation efficiency about computing the coefficient matrix.The storage is caused by improper storage of sparse coefficient matrix.In this paper we compress the sparse matrices by the compressed sparse row (CSR)format and em-ploy the following strategy:firstly,the value is computed at the corresponding model nodes within the influence domain of each Gauss point;secondly,the summation is performed within the influence domain of different Gauss points that contain the same model node.In the calculation of wave field time recursive,we solve the linear equations with the help of linear sparse solver'PARDISO'.In order to further improve the computation efficiency,we use OpenMP basing the existing FDM data.Through the above methods,we can effectively solve these problems of memory limit and computational efficiency.The final results in-dicate that the above methods are accurate and efficient.