The Application of CPU/GPU Collaborative Computing in the Frequency Domain 2D Full Waveform Inversion

The full waveform inversion(FWI) uses the kinematic and kinetic information of wave field to reconstruct the physical parameters underground, and it is an effective method to establish the high precision velocity model. A huge amount of calculation is one of the bottleneck of restricting its practicality. In this paper, we use the parallel strategy combined thick and thin to solve the complex computing problem of frequency domain FWI. To improve the computational efficiency of the FWI, the MPI technology was applied to compute multi shot forward parallel, while the GPU technology was used to speed up the solving of large sparse linear algebraic equations. The theoretical model verified the correctness and effectiveness of the method, and the correlation analysis results of different amount of data and GPU calculation efficiency are given. This paper proposed the bottleneck and the development direction of the CPU/GPU cooperative parallel computing in the frequency domain FWI.