Study Of 3D Finite-difference Transient Electromagnetic Forward Modeling With OpenACC

In recent years, with the continuous development of the transient electromagnetic technology and the progress of the transient electromagnetic instrument, we can collect more and more accurate and reliable transient electromagnetic field data. However, the study of transient electromagnetic method of our team is still not deep enough to play a good role. At present, the inversion of transient electromagnetic field is mainly based on one dimension. In order to further study the transient electromagnetic method, the finite difference time domain 3D transient electromagnetic forward in-depth study, but the routine calculation on CPU takes too long to model too long computing time to endure. With the development of GPU parallel computing in recent years, a large number of GPU parallel technologies including CUD A, OpenGL, OpenAcc have been formed. But because of the parallel technology of CUDA and OpenGL on the traditional hardware structure should be highly combined, so programming is difficult, and has written procedures for transplantation is poor, so the technology popularization and application is limited. OpenAcc technology has made a great progress in these aspects.In this paper,using a new type of GPU parallel programming technology of OpenAcc three-dimensional finite difference time domain transient electromagnetic forward algorithm, greatly shorten the computation time of 3D modeling, increase the calculation efficiency, and carried out the following studies on the:First of all, the application of Yee (1966) proposed a staggered grid scheme and Wang&Hohmann (1993) DuFort-Frankel scheme is proposed to push on uniform grid Maxwell equations based on the partition of the FDTD discrete form, and use soft source directly into the way of applying the excitation source, the deduced Max Well finite difference equations of active medium. A brief introduction on the development and advantage of GPU parallel technology, introduces the implementation mode and advantage of OpenAcc parallel technology, practical application examples show that the OpenAcc parallel technology compared to the traditional CUDA technology and the advantages of this paper is selected as research tools;Secondly, we analyze the 3D forward algorithm of FDTD, find out the most time-consuming part of the program, and use OpenAcc parallel technology to process it. The accuracy of the parallel program is verified by using the classical homogeneous half space model, the layered model and the 3D model;Finally, the speedup of the parallel program is evaluated, and the maximum speedup of the parallel program can reach 20x. Based on a large number of model tests, the upper bound of the parallel program in the computing environment is analyzed,and the influence of the mesh generation method on the speedup of parallel programs is analyzed. At last, it is proved that the accuracy of the parallel program is very high in most geological models and the acceleration performance is satisfactory.