基于Xeon Phi平台的波动方程叠前深度偏移  

Wave equation prestack depth migration based on Xeon Phi platform

作　　者：杨祥森[1] 金君[2] 王鹏[1] 马召贵[1] 亢永敢[1] 

机构地区：[1]中国石化石油物探技术研究院,江苏南京211113 [2]Intel亚太研发有限公司,上海200241

出　　处：《计算机工程与科学》2015年第5期907-913,共7页Computer Engineering & Science

摘　　要：波动方程叠前深度偏移适用于强横向变速介质,是一种高精度成像方法,但其巨大的计算量阻碍了该技术的应用。Xeon Phi是一种全新的高性能计算设备,为波动方程叠前深度偏移方法的推广应用提供了新的技术支持。以裂步傅里叶算子为例,介绍了面向Xeon Phi平台的偏移算法移植和优化方法,即采用offload模式将计算核函数加载到Xeon Phi设备上,在Xeon Phi协处理器上采用多线程方式,并且调整程序结构,充分利用SIMD矢量引擎提高向量化处理效率。扩展负载动态均衡的并行框架,形成了一套适用于大规模异构系统、基于Xeon Phi平台的波动方程叠前深度偏移软件。实际数据测试表明Xeon Phi平台可以极大地提高地震偏移处理效率,具有良好的可扩展性。Wave equation prestack depth migration is a high-precision seismic imaging method, suit- able for complex media with sharp lateral velocity variation, however, its huge computation hinders its application. Xeon Phi is a new product for throughput computing which provides a better choice for these methods. Taking the Split Step Fourier algorithm for example, we introduce the code migration and per- formance optimization on Xeon Phi platform. We offload the kernel which is fully threaded to Xeon Phi by predefined compiler pragmas so as to take full advantage of SIMD engine to improve the efficiency of the parallelization processing; and we change the code structure to ensure the vectorization. Besides, by extending the parallel processing framework with the dynamic workload-balance mechanism, we develop a seismic imaging system based on Xeon Phi platform which is suitable for large-scale, full hybrid hetero- geneous clusters. Finally, we conduct a real workload test and the results show that Xeon Phi platform can significantly improve the seismic imaging efficiency,and has good core scalability.

关 键 词：XEON Phi平台 波动方程叠前深度偏移 裂步傅里叶算子 并行框架 offload模式 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]