Accelerating 3D fourier migration on graphics processing units

Summary: Computational cost is a major factor that inhibits the practical application of 3D depth migration. We present a fast parallel scheme to speed up 3D wave-equation depth migration on a parallel computing device, Graphics Processing Units (GPUs). The third-order generalizedscreen propagator is employed to take advantage of the built-in software implementation of fast Fourier transform. We minimize the low-bandwidth data transfer by storing the 3D velocity model and imaged data in device memory. We reduce half of the memory demand by compressing the 3D velocity model and imaged data using integer arrays instead of float arrays. A 2D tapered function is suggested for boundary conditions, which is more suitable for the parallel implementation on the GPU than the commonlyused 1D one. Both three-dimensional impulse responses and synthetic data examples demonstrate that the proposed GPU-based Fourier migration is typically 25 to 40 times faster than the CPU-based implementation. It enables us to build the macro velocity model in a much shorter turnaround time.