CPU与GPU协同并行的多分量地震数据各向异性叠前时间偏移

doi:10.13810/j.cnki.issn.1000-7210.2019.01.008

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (4482 KB) HTML []
输出: BibTeX | EndNote (RIS)

摘要地震勘探工区规模的日益庞大，造成多分量地震数据的各向异性叠前时间偏移算法耗时巨大。目前常用CPU集群方式并行加速该类算法，而集群方法必将导致节点间通信时耗增大；同时受限于CPU结构特点，只能通过扩大集群规模提高加速比。针对上述问题，提出一种基于CPU与GPU协同并行的多分量地震数据各向异性叠前时间偏移算法，利用OpenMP和CUDA实现CPU与多个GPU的协同并行，使用内存映射方法降低I/O耗费；并根据CPU与GPU的结构特点优化地震数据及速度数据的读取和存储方式，分割成像空间以节省算法内存消耗，每次只计算与开启的GPU个数相同条数的主测线，主测线内部采用一个GPU线程对应一道地震数据的偏移计算方法，以充分利用GPU计算能力。应用约29G的实际工区多分量地震数据分别比较不同个数GPU协同CPU并行的加速比，得知使用6个GPU协同CPU对实际纵波及转换波数据进行并行偏移处理时，加速比分别达到444和449。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	刘帅
	季晓慧
	芦俊
	荣骏召

关键词 ：多分量地震数据, 各向异性, 叠前时间偏移, CUDA, OpenMP

Abstract：The anisotropy prestack time migration (PSTM) algorithm costs a lot of time in practical application due to current seismic survey sizes.CPU-cluster computing is usually used to accelerate PSTM,but it leads to the increase of communication cost between the computing nodes.According to the CPU structure,the only way to improve a cluster performance is to enlarge the scale of cluster.To solve this problem,a multi-component anisotropy PSTM algorithm based on collaborative parallel computing with CPU and GPU is proposed.This algorithm implements CPU collaborative parallel computing with a multi-GPUs environment and OpenMP and compute unified device architecture (CUDA).It decreases I/O costs with the memory map method,and optimizes reading and storing seismic data according to the structural features of CPU and GPU.The algorithm segments image volume to decrease memory occupation,computes the number of inlines which is equal to the number of available GPUs.One GPU thread will image one seismic data trace amongst inline.The propose approach was tested on 29GB seismic data with different numbers of GPUs.Between P-P wave and P-S wave seismic data,speedups are 444 and 449 when parallel computing with CPU collaborative parallel computing to 6 GPUs.

Key words： multi-component data anisotropy pre-stack time migration compute unified device architecture (CUDA) OpenMP

收稿日期: 2018-04-30

PACS:

P631

基金资助:本项研究得到国家自然科学基金项目“基于异构多核并行计算的地下水数值模拟求解方法及应用”（41102158）、“薄互层多波AVO响应与反演方法研究”（41574126）、中央高校基本科研业务费课题、中国石油天然气集团公司“弹性波地震成像技术合作研发”项目等联合资助。

通讯作者: 季晓慧,北京市海淀区学院路29号中国地质大学(北京)信息工程学院,100083。Email:xhji@cugb.edu.cn E-mail: xhji@cugb.edu.cn

作者简介: 刘帅,硕士研究生,1994年生;2015年获中国地质大学(北京)计算机科学与技术专业学士学位;2015年被中国地质大学(北京)信息工程学院录取为计算机科学与技术专业硕士研究生;主要从事高性能计算方面的学习和研究。

引用本文:

刘帅, 季晓慧, 芦俊, 荣骏召. CPU与GPU协同并行的多分量地震数据各向异性叠前时间偏移[J]. 石油地球物理勘探, 2019, 54(1): 65-72. LIU Shuai, JI Xiaohui, LU Jun, RONG Junzhao. Multi-component anisotropy prestack time migration based on collaborative parallel computing with CPU and GPU. Oil Geophysical Prospecting, 2019, 54(1): 65-72.

链接本文:

http://www.ogp-cn.com/CN/10.13810/j.cnki.issn.1000-7210.2019.01.008 或 http://www.ogp-cn.com/CN/Y2019/V54/I1/65

[1]	赵波,王赟,芦俊.多分量地震勘探技术新进展及关键问题探讨[J].石油地球物理勘探,2012,47(3):506-516.ZHAO Bo,WANG Yun and LU Jun.Recent advances of multi-component seismic and some of its key issues[J].Oil Geophysical Prospecting,2012,47(3):506-516.
[2]	周丽,顾汉明,成景旺,等.基于MPI的OBC三维多波多分量地震观测正演模拟并行算法实现[J].石油物探,2014,53(6):665-674.ZHOU Li,GU Hanming,CHENG Jingwang,et al.Algorithm realization of parallel forward modeling of 3D multi-wave and multicomponent seismic survey for OBC observation based on MPI[J].Geophysical Prospecting for Petroleum,2014,53(6):665-674.
[3]	荣骏召,芦俊,李建峰,等.矢量Kirchhoff叠前深度偏移[J].石油地球物理勘探,2017,52(6):1170-1176.RONG Junzhao,LU Jun,LI Jianfeng,et al.Vector prestack depth migration based on Kirchhoff integral equation[J].Oil Geophysical Prospecting,2017,52(6):1170-1176.
[4]	赵长海,罗国安,张旭东,等.大规模异构集群上Kirchhoff叠前时间偏移并行算法[J].石油地球物理勘探,2016,51(5):1040-1048.ZHAO Changhai,LUO Guoan,ZHANG Xudong,et al.Kirchhoff prestack time migration on large heterogeneous computing systems[J].Oil Geophysical Prospecting,2016,51(5):1040-1048.
[5]	马召贵,赵改善,武港山,等.基于CPU+GPU联合计算真地表叠前时间偏移实用化研究[J].石油地球物理勘探,2014,49(3):530-537.MA Zhaogui,ZHAO Gaishan,WU Gangshan,et al.Kirchhoff pre-stack time migration of complex surface data on CPU+GPU[J].Oil Geophysical Prospecting,2014,49(3):530-537.
[6]	熊敏,王勇献.面向CPU+MIC混合易购平台的地震波叠前时间偏移算法并行与优化策略[J].计算机工程与科学,2015,37(1):13-22.XIONG Min,WANG Yongxian.Parallel optimization of the seismic wave PKTM algorithm on CPU+MIC heterogeneous platform[J].Computer Engineering & Science,2015,37(1):13-22.
[7]	蔡志成,顾汉明,成景旺,等.大规模变网格三维地震正演MPI并行策略与实现[J].石油地球物理勘探,2017,52(3):468-476.CAI Zhicheng,GU Hanming,CHENG Jingwang,et al.Parallel large-scale variable grid 3D seismic wave modeling[J].Oil Geophysical Prospecting,2017,52(3):468-476.
[8]	马召贵,赵改善,武港山,等.起伏地表叠前时间偏移的多级并行优化技术[J].石油物探,2013,52(3):280-287.MA Zhaogui,ZHAO Gaishan,WU Gangshan,et al.Multi-level parallel optimization technique for pre-stack time migration from rugged topography[J].Geophysical Prospecting for Petroleum,2013,52(3):280-287.
[9]	Alves M M,Pestana R D C,Silva R A P D,et al.Accelerating pre-stack Kirchhoff time migration by manual vectorization[J].Concurrency & Computation Practice & Experience,2016.
[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 and ZHOU Shiyong.Implement Kirchhoff prestack time migration algorithm on CUDA architecture[J].Application Research of Computers,2009,26(12):4474-4477.
[11]	喻勤,张少华,孔选林.基于MPI和CUDA的转换波Kirchhoff叠前时间偏移并行计算[J].石油物探,2013,52(1):60-64.YU Qin,ZHANG Shaohua and KONG Xuanlin.Converted-wave Kirchhoff prestack time migration parallel computation based on MPI+CUDA[J].Geophy-sical Prospecting for Petroleum,2013,52(1):60-64.
[12]	亢永敢,赵改善,魏嘉,等.基于Hadoop的Kirchhoff叠前时间偏移并行算法[J].石油地球物理勘探,2015,50(6):1213-1218.KANG Yonggan,ZHAO Gaishan,WEI Jia,et al.Para-llel algorithms of Kirchhoff prestack time migration based on Hadoop[J].Oil Geophysical Prospecting,2015,50(6):1213-1218.
[13]	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 moutainous regions and its GPU implementation[J].Chinese Journal of Geophysics,2009,52(21):3101-3108.
[14]	杨震,芦俊,王赟,等.转换波时距曲线精度对比[C].中国地球科学联合学术年会,北京,2015.
[15]	Alkhalifah T,Tsvankin I.Velocity analysis for transversely isotropic media[J].Geophysics,1995,60(5):1550-1566.
[16]	Dai H.Parallel processing of prestack Kirchhoff time migration on a PC cluster[J].Computers & Geosciences,2005,31(7):891-899.
[17]	Yao L,Ji X,Liu S,et al.Parallel implementation and performance comparison of BICGStab for massive sparse linear system of equations on GPU libraries[C].The 12th Annual International Conference on Advanced and Trusted Computing(UIC-ATC-ScalCom-CBDCom-IoP),Beijing,China,2015,603-608.
[18]	Panetta J,Teixeira T,de Souza Filho P R P,et al.Accelerating time and depth seismic migration by CPU and GPU cooperation[J].International Journal of Parallel Programming,2012,40(3):290-312.
[19]	Shi X,Li C,Wang S.Computing prestack Kirchhoff time migration on general purpose[J].Computers & Geosciences,2011,37(10):1702-1710.
[20]	Rahmad M H,Meng S S,Karuppiah Ek,et al.Comparision of CPU and GPU implementation of computing absolute difference[C].International Conference on Control System,Computing and Engineering(ICCSCE),IEEE,Penang,Malaysia,2011,132-137.
[21]	白洪涛,范翠松,李昂,等.基于OpenMP+MPI的2.5维电磁场正演并行化方法[J].石油地球物理勘探,2015,50(1):170-176.BAI Hongtao,FAN Cuisong,LI Ang,et al.2.5 dimension eletromagnetic forward parallel computing based on MPI+OpenMP[J].Oil Geophysical Prospecting,2015,50(1):170-176.
[22]	Tian Xinmin,Bik Aart,Girkai Milind,et al.Intel OpenMP C++/Fortran compiler for hyper-threading technology:Implementation and Performance[J].Intel Technology Journal,2002:6(1):1-11.
[23]	王玉凤.三维多波地震资料逆时偏移成像技术及GPU并行实现[D].山东青岛:中国海洋大学,2014.WANG Yufeng.Reverse-time Migration Imaging and GPU Parallel Implementation of 3D Multi-wave Seismic Data[D].Ocean University of China,Qingdao,Shandong,2014.
[24]	喻勤,程冰洁,孔选林.基于CUDA的转换波Kirchhoff叠前时间偏移算法研究与实现[J].石油地球物理勘探,2013,48(1):58-63.YU Qin,CHENG Bingjie and KONG Xuanlin.Converted-wave Kirchhoff prestack time migration based on CUDA[J].Oil Geophysical Prospecting,2013,48(1):58-63.