Method of Prestack Time Migration of Seismic Data of Mountainous Regions and Its GPU Implementation

Abstract

The geological structure in the mountainous regions of western China is quite complicated, and the elevation varies violently. At present, the main approach for seismic data imaging is the prestack Kirchhoff time migration, however, violent variation of surface elevation makes the datum plane selection of prestack time migration very difficult. This article proposed a new prestack time migration method through modifying the routine travel time computation, this method can greatly raise the imaging effect in mountainous region with obvious variation of surface elevation. In addition, this paper also introduced application of GPU in prestack time migration, by which the prestack time migration has been optimized and implemented, and the computation efficiency of single NVIDIA Tesla C1060 GPU can be more than 70 times faster than single core CPU with main frequency of 2.5 GHZ when without the I/O problem.

Practical Implementation of Prestack Kirchhoff Time Migration on a General Purpose Graphics Processing Unit

“…Some studies have also used NVIDIA GPUs to accelerate PSTM. In particular, Liu et al (2009) discussed the possibility of parallel computation with NVIDIA GPUs. Shi et al (2011) proposed a method for accelerating PSTM on GPUs by splitting the PSTM procedure into four consequence kernels according to the GPU memory limitations, as well as considering the floating point error problem, which may lead to differences when comparing PSTM with CPU computations.…”

Computing Prestack Kirchhoff time migration algorithm on OpenCL and GPGPU

“…PKTM could take days or more to generate the final migrated images for some seismic data on high-performance clusters with hundreds of CPUs, So how to implement Kirchhoff algorithm on generous purpose GPUs is one of the research hotspots in parallel computing and geophysical exploration areas. optimization algorithm on GPGPU for seismic data processing of existing and without simplifying or redesigning the mathematical operations have been discussed [6] [7].GPGPU and CUDA as accelerating platforms in migrating software have been evaluated [8], a new Kirchhoff migration method for GPGPU have been proposed [9] [10]. However, there is little published in the literature in terms of how to implement and optimize Kirchhoff algorithm on OpenCL.…”

3D gravity interface inversion constrained by a few points and its GPU acceleration

“…We can found detailed program guide and SDKs on GPUs with CUDA in the literature of Sanders NVIDIA Corporation (2010). With CUDA, GPU parallel computing has been widely applied in gravity modelling, seismic processing, astronomy, molecular dynamics simulation, fluid mechanics and some other fields (Chen,2012,Cuma et al, 2012, Hegeman, 2006Stone, 2007;Tolke, 2008;Belleman,2008;Gaburov,2009;Liu et al, 2009;Wong et al, 2011;Martin et al, 2013). This paper has two objectives.…”