应用图形处理器快速计算逆时偏移
|
摘要
逆时偏移是目前精度最高的地震数据叠前深度偏移方法,但高强度的计算需求限制了其在工业生产领域的大规模应用。可编程图形处理器的发展为逆时偏移的快速计算提供了一种新的计算选择。围绕如何在图形处理器上开展逆时偏移计算展开,总结了图形处理器计算的优化关键,并根据逆时偏移的特点着重介绍了两个优化环节:一个是应用随机边界条件,以计算换存储,减少数据在主机和图形处理器间的传输;二是应用共享存储器来存储正演计算的波场,相比全局存储器,提高了数据读取的带宽。应用Marmousi模型数据对经过上述优化后的程序进行了测试,结果表明,图形处理器逆时偏移程序得到了很好的优化,提高了计算效率。
Reverse-time migration is the most accurate seismic data prestack depth migration method,but the computing needs of its high-intensity limit its large-scale computing applications.The development of programmable Graphics Processing Unit(GPU) provides an alternative method for rapid calculation of the reverse-time migration.In this paper,we focus on the main steps using GPU to calculate the reverse time migration.We summarize the key points in GPU code optimization,and highlight two optimized components: one is the application of random boundary conditions.The data transmission between the host and GPU is reduced.The second is the application of shared memory to store the forward wave field.Compared to the global memory,data reading band widths increase greatly.Finally,we use the Marmousi data to test our code,and the results show that the reverse-time migration program running on GPU has been well optimized to improve the computational efficiency.
Reverse-time migration is the most accurate seismic data prestack depth migration method,but the computing needs of its high-intensity limit its large-scale computing applications.The development of programmable Graphics Processing Unit(GPU) provides an alternative method for rapid calculation of the reverse-time migration.In this paper,we focus on the main steps using GPU to calculate the reverse time migration.We summarize the key points in GPU code optimization,and highlight two optimized components: one is the application of random boundary conditions.The data transmission between the host and GPU is reduced.The second is the application of shared memory to store the forward wave field.Compared to the global memory,data reading band widths increase greatly.Finally,we use the Marmousi data to test our code,and the results show that the reverse-time migration program running on GPU has been well optimized to improve the computational efficiency.
引文
[1]Bednar J B,Bednar C J.Two-way versus one-way:A case studystyle comparison[M]//Society of Exploration Geophysicists.76th Annual International Meeting,SEG Expanded Abstracts.Houston:SEG,2006:2343-2347.
[2]Hemon C.Equations d'onde et modeles[J].Geophysical Pros-pecting,1978,26(1):790-821.
[3]Baysal E,Kosloff D D,Sherwood J W C.Reverse time migration[J].Geophysics,1983,48(2):1514-1524.
[4]McMechan G A.Migration by extrapolation of time-dependentboundary values[J].Geophysical Prospecting,1983,31(4):413-420.
[5]Loewenthal D,Mufti I R.Reversed time migration in spatial fre-quency domain[J].Geophysics,1983,48(3):627-635.
[6]Farmer P A,Jones I F,Zhou H,et al.Application of reversetime migration to complex imaging problems[J].First Break,2006,24:65-74.
[7]Micikevicius P.3d finite difference computation on GPUS usingCUDA[M]//ACM.Proceedings of 2nd Workshop on GeneralPurpose Processing on Graphics Processing Units.Washington DC:ACM,2009:79-84.
[8]Yoon K,Shin C,Suh S,et al.3D reverse-time migration usingthe acoustic wave equation:An experience with the SEG/EAGEdata set[J].The Leading Edge,2003,22:38-41.
[9]Nemeth T,Stefani J,Liu W,et al.An implementation of theacoustic wave equation on FPGAS[M]//Society of ExplorationGeophysicists.78th Annual International Meeting,SEG Expand-ed Abstracts.Houston:SEG,2008:2874-2877.
[10]Youngseo K,Dong-Joo M,Changsoo S.Frequency-domain re-verse-time migration with source estimation[J].Geophysics,2011,76(2):S41-S49.
[11]Xinyi Sun,Sang Suh.Maximizing Throughput for High Perform-ance TTI-RTM:From CPU-RTM to GPU-RTM[M]//Societyof Exploration Geophysicists.81th Annual International Meeting,SEG Expanded Abstracts.Houston:SEG,2011:10-20.
[12]NVIDIA Corporation.NVIDIA GPU computing developer homepage[EB/OL].http://developer.nvidia.com/object/gpucom-puting.html,2011-02-23/2012-01-10.
[13]NVIDIA Corporation.CUDA C programming best practices guides3.1[M/OL].http://www.nvidia.com/object/cuda_develop.html,2009-07-21/2012-03-29.
[14]Wang X,Gao X,Yao Z X.Accelerating POCS interpolation of3D irregular seismic data with Graphics Processing Units[J].Computers and Geosciences,2010,36:1292-1300.
[15]刘国峰,刘洪,李博,等.山地地震资料逆时偏移方法及其GPU实现[J].地球物理学报,2009,52(12):3101-3108.
[16]石颖,刘洪.基于GPU的表面多次波预测[J].石油地球物理勘探,2010,54(4):540-544.
[17]牟永光,裴正林.三维复杂介质地震波数值模拟[M].北京:石油工业出版社,2005:10-20.
[18]Symes W M.Reverse time migration with optimal checkpointing[J].Geophysics,2007,72(5):SM213-SM221.
[19]Clapp R G.Reverse time migration with random boundaries[M]//Society of Exploration Geophysicists.79th Annual Inter-national Meeting,SEG Expanded Abstracts.Houston:SEG,2009:2809-2813.
[2]Hemon C.Equations d'onde et modeles[J].Geophysical Pros-pecting,1978,26(1):790-821.
[3]Baysal E,Kosloff D D,Sherwood J W C.Reverse time migration[J].Geophysics,1983,48(2):1514-1524.
[4]McMechan G A.Migration by extrapolation of time-dependentboundary values[J].Geophysical Prospecting,1983,31(4):413-420.
[5]Loewenthal D,Mufti I R.Reversed time migration in spatial fre-quency domain[J].Geophysics,1983,48(3):627-635.
[6]Farmer P A,Jones I F,Zhou H,et al.Application of reversetime migration to complex imaging problems[J].First Break,2006,24:65-74.
[7]Micikevicius P.3d finite difference computation on GPUS usingCUDA[M]//ACM.Proceedings of 2nd Workshop on GeneralPurpose Processing on Graphics Processing Units.Washington DC:ACM,2009:79-84.
[8]Yoon K,Shin C,Suh S,et al.3D reverse-time migration usingthe acoustic wave equation:An experience with the SEG/EAGEdata set[J].The Leading Edge,2003,22:38-41.
[9]Nemeth T,Stefani J,Liu W,et al.An implementation of theacoustic wave equation on FPGAS[M]//Society of ExplorationGeophysicists.78th Annual International Meeting,SEG Expand-ed Abstracts.Houston:SEG,2008:2874-2877.
[10]Youngseo K,Dong-Joo M,Changsoo S.Frequency-domain re-verse-time migration with source estimation[J].Geophysics,2011,76(2):S41-S49.
[11]Xinyi Sun,Sang Suh.Maximizing Throughput for High Perform-ance TTI-RTM:From CPU-RTM to GPU-RTM[M]//Societyof Exploration Geophysicists.81th Annual International Meeting,SEG Expanded Abstracts.Houston:SEG,2011:10-20.
[12]NVIDIA Corporation.NVIDIA GPU computing developer homepage[EB/OL].http://developer.nvidia.com/object/gpucom-puting.html,2011-02-23/2012-01-10.
[13]NVIDIA Corporation.CUDA C programming best practices guides3.1[M/OL].http://www.nvidia.com/object/cuda_develop.html,2009-07-21/2012-03-29.
[14]Wang X,Gao X,Yao Z X.Accelerating POCS interpolation of3D irregular seismic data with Graphics Processing Units[J].Computers and Geosciences,2010,36:1292-1300.
[15]刘国峰,刘洪,李博,等.山地地震资料逆时偏移方法及其GPU实现[J].地球物理学报,2009,52(12):3101-3108.
[16]石颖,刘洪.基于GPU的表面多次波预测[J].石油地球物理勘探,2010,54(4):540-544.
[17]牟永光,裴正林.三维复杂介质地震波数值模拟[M].北京:石油工业出版社,2005:10-20.
[18]Symes W M.Reverse time migration with optimal checkpointing[J].Geophysics,2007,72(5):SM213-SM221.
[19]Clapp R G.Reverse time migration with random boundaries[M]//Society of Exploration Geophysicists.79th Annual Inter-national Meeting,SEG Expanded Abstracts.Houston:SEG,2009:2809-2813.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心