基于GPU并行加速的叠前逆时偏移方法

为了提高复杂地下介质的成像精度和偏移算法的计算效率,提出可高效对地下复杂构造进行准确成像的GPU加速叠前逆时偏移方法.该方法采用双程声波方程进行波场延拓,突破倾角限制,借助于高阶有限差分方法实现叠前逆时偏移成像;利用GPU(Graphic Processing Unit)并行加速技术对波场延拓和成像进行计算,相比于传统算法,其计算效率有较大提高,可以解决叠前逆时偏移算法计算量过大问题;在获取波场信息过程中,也采用随机边界条件,实施以计算换存储策略,解决逆时偏移计算中的海量存储问题.模型测试结果表明,该方法能够高效和高精度地对地下复杂地质体成像.

Prestack reverse time migration based on GPU parallel accelerating algorithm

In order to improve the complex subsurface imaging accuracy and computational efficiency of the algorithm,this paper presents an algorithm of prestack reverse time migration based on GPU(Graphic Processing Unit) accelerating which can image the underground complex structure effectively and accurately.By two-way wave equation to calculate wave field extrapolation,prestack reverse-time migration can overcome the dip limit,and the imaging algorithm is performed by high order finite difference in the paper.Wave field extrapolation and imaging condition are calculated by GPU parallel accelerating technology,comparing to conventional algorithm,its computation efficiency has been greatly improved,and it meets large amount of computation requirement in prestack reverse-time migration.The random boundary condition approach is adopted to obtain wavefield information,which reduces the memory demand but sacrifices the computation cost,and it solves the massy memory problem in reverse time migration.The tests on model illustrate that this approach can imaging complicated geological body efficiently and precisely.