Kirchhoff叠前时间偏移的GPU移植与性能优化技术

doi:10.7623/syxb201404010

石油学报 ›› 2014, Vol. 35 ›› Issue (4): 700-705.DOI: 10.7623/syxb201404010

Kirchhoff叠前时间偏移的GPU移植与性能优化技术

马召贵^1,2, 赵改善¹, 武港山², 岳承琪¹, 何恺¹, 王鹏¹

1. 中国石油化工股份有限公司石油物探技术研究院江苏南京 211103;
2. 南京大学软件新技术国家重点实验室江苏南京 210093

收稿日期:2014-01-17 修回日期:2014-04-15 出版日期:2014-07-25 发布日期:2014-08-05
通讯作者: 马召贵，男，1978年12月生，2003年获石油大学（北京）学士学位，2010年获中国石油大学（北京）博士学位，现为中国石油化工股份有限公司石油物探技术研究院高级工程师，主要从事地震偏移成像方法研究与软件研发工作。Email：mazhg.swty@sinopec.com
作者简介:马召贵，男，1978年12月生，2003年获石油大学（北京）学士学位，2010年获中国石油大学（北京）博士学位，现为中国石油化工股份有限公司石油物探技术研究院高级工程师，主要从事地震偏移成像方法研究与软件研发工作。Email：mazhg.swty@sinopec.com
基金资助:
国家高技术研究发展计划（863）项目（2009AA01A140）资助。

GPU-based porting and optimization of Kirchhoff pre-stack time migration

Ma Zhaogui^1,2, Zhao Gaishan¹, Wu Gangshan², Yue Chengqi¹, He Kai¹, Wang Peng¹

1. Sinopec Geophysical Research Institute, Jiangsu Nanjing 211103, China;
2. State Key Laboratory for Novel Software Technology, Nanjing University, Jiangsu Nanjing 210093, China

Received:2014-01-17 Revised:2014-04-15 Online:2014-07-25 Published:2014-08-05

摘要/Abstract

摘要：

叠前时间偏移在工业生产中发挥着极其重要的作用，为了提高该算法的计算效率，开展了基于GPU异构计算平台的算法移植与优化。首先根据积分法偏移的算法特点制定了偏移距域的多进程数据域并行以及IO与计算异步并行总体并行策略；然后为了提高偏移核心计算部分在GPU上的计算效率，对偏移计算核在GPU上的并行方案进行了分析，选择了成像域超大规模线程并行方案对算法进行了移植和优化，并对不同优化手段在不同GPU硬件平台下获得的性能加速进行了对比测试；最后利用大规模计算节点及大规模地震数据体进行了移植后算法的应用测试，并对算法的计算效率、可扩展性以及精度误差进行了分析。大规模应用测试表明，积分法叠前时间偏移经过GPU移植后可获得较CPU平台近7倍的性能提升，具有很好的工业应用价值。

关键词: 叠前时间偏移, 高性能计算, 并行效率, GPU移植, CUDA优化

Abstract:

Pre-stack time migration (PSTM) plays an important role in industrial production. To improve its computation efficiency, algorithm porting and optimization are performed based on the GPU heterogeneous computing platform. According to the characteristics of the Kirchhoff algorithm, a parallel strategy is initially designed with the multi-processes parallelism in common offset domain and the asynchronous parallelism of IO and computing. For improving the computation efficiency of migration kernel on GPU, this paper then analyzes the parallel strategy on GPU of migration kernel, performs algorithm porting and optimization with massive scale threads in imaging domain, and assesses the performance of different optimization methods for various GPU hardware platforms. Finally, the algorithm testing is carried out with large-scale computing nodes and seismic data followed by an analysis of computation efficiency, scalability, and precision error of the algorithm. Large-scale application test results show that after GPU porting, the performance of Kirchhoff pre-stack time migration is improved by nearly 7-fold compared with the CPU platform, thus having great value for industrial applications.

Key words: pre-stack time migration, high performance computing, parallel efficiency, GPU porting, CUDA optimization

中图分类号:

P631.443

马召贵, 赵改善, 武港山, 岳承琪, 何恺, 王鹏. Kirchhoff叠前时间偏移的GPU移植与性能优化技术[J]. 石油学报, 2014, 35(4): 700-705.

Ma Zhaogui, Zhao Gaishan, Wu Gangshan, Yue Chengqi, He Kai, Wang Peng. GPU-based porting and optimization of Kirchhoff pre-stack time migration[J]. ACTA PETROLEI SINICA, 2014, 35(4): 700-705.

参考文献

[1] 夏义平,徐礼贵,温铁民,等.叠前偏移及储层预测技术研发进展与应用[J].石油学报,2012,33(增刊1):132-141.
Xia Yiping,Xu Ligui,Wen Tiemin,et al.Progress and applications of prestack migration and reservoir forecasting technology[J].Acta Petrolei Sinica,2012,33(Supplement 1):132-141.
[2] 王宏琳.油气勘探开发计算机软件集成平台的研究[J].石油学报,1999,20(5):20-23.
Wang Honglin.A study on petroleum exploration and production software intergration platform[J].Acta Petrolei Sinica, 1999,20(5):20-23.
[3] 赵改善.我们需要多大和多快的计算机[J].勘探地球物理进展,2004,27(1):22-28.
Zhao Gaishan.How big and fast computers can meet our needs?[J].Progress in Exploration Geophysics,2004,27(1):22-28.
[4] 齐俊宁,朱敏,吴彦伟,等.海量数据叠前时间偏移方法[J].天然气工业,2008,28(11):46-48.
Qi Junning,Zhu Min,Wu Yanwei,et al.The method of prestack time migration in processing enormous data[J].Natural Gas Industry,2008,28(11):46-48.
[5] 李伟,顾乃杰,刘振宽.三维叠前Kirchhoff深度偏移软件在并行计算机上的实现技术[J].计算机工程与应用, 2002,20(4):211-214.
Li Wei,Gu Naijie,Liu Zhenkuan.3-D pre-stack Kirchhoff depth migration and its parallel calculation technique[J].Computer Engineering and Applications,2002,20(4):211-214.
[6] 李博,刘国峰,刘洪.地震叠前时间偏移的一种图形处理器提速实现方法[J].地球物理学报,2009,52(1):245-252.
Li Bo,Liu Guofeng,Liu Hong.A method of using GPU to accelerate seismic pre-stack time migration[J].Chinese Journal of Geophysics,2009,52(1):245-252.
[7] 董春晖,张剑锋.起伏地表下的直接叠前时间偏移[J].地球物理学报,2009,52(1):239-244.
Dong Chunhui,Zhang Jianfeng.Prestack time migration including surface topography[J].Chinese Journal of Geophysics,2009,52(1):239-244.
[8] 刘国峰,刘洪,李博,等.山地地震资料叠前时间偏移方法及其GPU实现[J].地球物理学报,2009,52(12):3101-3108.
Liu Guofeng,Liu Hong,Li Bo,et al.Method of prestack time migration of seismic data of mountainous regions and its GPU implementation[J].Chinese Journal of Geophysics,2009,52(12):3101-3108.
[9] 刘国峰,刘钦,李博,等.油气勘探地震资料处理GPU/CPU协同并行计算[J].地球物理学进展,2009,24(5):1671-1678.
Liu Guofeng,Liu Qin,Li Bo,et al.GPU/CPU co-processing parallel computation for seismic data processing in oil and gas exploration[J].Progress in Geophysics,2009,24(5):1671-1678.
[10] 李肯立,彭俊杰,周仕勇.基于CUDA的Kirchhoff叠前时间偏移算法设计与实现[J].计算机应用研究,2009,26(12):4474-4477.
Li Kenli,Peng Junjie,Zhou Shiyong.Implement Kirchhoff prestack time migration algorithm on CUDA architecture[J].Application Research of Computers,2009,26(12):4474-4477.
[11] He Chuan,Sun Chuanwen,Lu Mi,et al.Prestack Kirchhoff time migration on high performance reconfigurable computing platform [C].Proceedings of SEG 2005 International Exposition and 75th Annual Meeting,Houston,November,6-11,2005:1902-1905.
[12] Panetta J,Teixeira T,Souza Filho P,et al.Accelerating Kirchhoff Migration by CPU and GPU Cooperation[C].Proceedings of the 21st International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD),IEEE Computer Society,2009:26-32.
[13] Panetta J.Seismic imaging on novel computer architectures[C].Proceedings of the 11th International Congress of the Brazilian Geophysical Society & EXPOGEF 2009,Saluador,Bahia,Brazil,August 24-28,2009.
[14] Brouwer W,Natoli V.A novel GPGPU approach to Kirchhoff time migration[G].Expanded Abstracts of 81^th Annual Internat SEG Mtg,2011:3465-3469.
[15] Teixeira D,Yeh A,Gajawada S.Implementation of Kirchhoff prestack depth migration on GPU[G].Expanded Abstracts of 83^rd Annual Internat SEG Mtg,2013:3683-3686.
[16] 王华忠,蔡杰雄,孔祥宁,等.适于大规模数据的三维Kirchhoff积分法体偏移实现方案[J].地球物理学报, 2010,53(7):1699-1709.
Wang Huazhong,Cai Jiexiong,Kong Xiangning,et al.An implementation of Kirchhoff integral prestack migration for large-scale data[J].Chinese Journal of Geophysics,2010,53(7):1699-1709.
[17] 王华忠,刘少勇,孔祥宁,等.大规模三维地震数据Kirchhoff叠前深度偏移及其并行实现[J].石油地球物理勘探,2012,47(3):404-410.
Wang Huazhong,Liu Shaoyong,Kong Xiangning,et al.3D Kirchhoff PSDM for large-scale seismic data and its parallel implementation strategy[J].Oil Geophysical Prospecting,2012,47(3):404-410.

Kirchhoff叠前时间偏移的GPU移植与性能优化技术

GPU-based porting and optimization of Kirchhoff pre-stack time migration

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 3

编辑推荐

Metrics

本文评价

[1]	金龙, 陈小宏, 姜香云. 利用地震资料定量反演孔隙度和饱和度的新方法[J]. 石油学报, 2006, 27(4): 63-66.
[2]	刘洋. 利用地震资料估算孔隙度和饱和度的一种新方法[J]. 石油学报, 2005, 26(2): 61-64.
[3]	宋维琪, 赵万金, 冯磊, 吴华. 地震高分辨率反演和地质模拟联合预测薄储层[J]. 石油学报, 2005, 26(1): 50-54.