【摘要】 转换波叠前时间偏移计算量巨大、耗费时间长,影响了多波多分量地震数据的处理效率,也限制了转换波技术在生产上的应用规模。目前转换波叠前时间偏移主要采取CPU集群计算方式,但CPU集群存在功耗大、占用空间大和维护成本高等缺点,为缩短偏移计算耗时和降低计算成本,本文提出一种基于CUDA技术的转换波Kirchhoff叠前时间偏移并行算法。应用理论数据和实际转换波数据在CPU和GPU测试平台对算法进行了对比验证和性能分析,确认GPU并行算法计算速度可提高上百倍,是一种高效、低成本的方法。更多还原

【Abstract】 Because of large simulation calculation and time consumption,the process of converted-wave(C-wave)Kirchhoff prestack time migration is an low-efficient data processing method and its application is restricted.CPU-cluster computing is applied widely at present;however CPU-clusters have some problem such as high power consumption,large space occupation,and high maintenance costs.For improving the efficiency and reducing the computing cost,we propose in this article a parallel algorithm of C-wave Kirchhoff prestack time migration based on CUDA.Theoretical and real C-wave data tests on the GPU and CPU platforms show that this algorithm is a high performance and low cost C-wave Kirchhoff prestack time migration method as it can improve the calculation efficiency by 100times.更多还原