一个基于多GPU的TTI介质逆时偏移方法（英文）
|
摘要
地震数据TTI介质逆时偏移计算量巨大,阻碍了其深度应用。NVIDIA? GPU以及其通用计算具有高度并行特性,为加速该类高密度计算提速提供了可能,可解决由于地震处理成像空间增加所带来的计算量大难题。本文提出了一种高效率的多GPU并行计算策略,以解决TTI介质逆时偏移庞大的数据处理问题。该策略流程,以GPU及其统一设备架构CUDA为前提,拥有多核心同时计算、作业流、点对点GPU直接传输等一些列运算特点。其核心是将GPU间边界数据的数据传输时间和不同区域的差分计算时间重叠。由于逆时偏移计算强度主要与差分计算空间相关。重叠计算后,数据传输时间可以忽略,因而计算效率随GPU数量的增加呈线性提高。并用于TTI逆时偏移成像的处理,以验证本文提出的高计算效率的多GPU并行计算策略的正确性。对比试验表明,利用本文提出的多GPU策略可大大提供高密度数据成像计算的效率及实现多GPU计算时效率呈线性增加,提升了该计算的延展特性。
Reverse time migration(RTM) is an indispensable but computationally intensive seismic exploration technique. Graphics processing units(GPUs) by NVIDIA? offer the option for parallel computations and speed improvements in such high-density processes. With increasing seismic imaging space, the problems associated with multi-GPU techniques need to be addressed. We propose an efficient scheme for multi-GPU programming based on the features of the compute-unified device Architecture(CUDA) using GPU hardware, including concurrent kernel execution, CUDA streams, and peer-to-peer(P2P) communication between the different GPUs. In addition, by adjusting the computing time for imaging during RTM, the data communication times between GPUs become negligible. This means that the overall computation efficiency improves linearly, as the number of GPUs increases. We introduce the multi-GPU scheme by using the acoustic wave propagation and then describe the implementation of RTM in tilted transversely isotropic(TTI) media. Next, we compare the multi-GPU and the unified memory schemes. The results suggest that the proposed multiGPU scheme is superior and, with increasing number of GPUs, the computational efficiency improves linearly.
Reverse time migration(RTM) is an indispensable but computationally intensive seismic exploration technique. Graphics processing units(GPUs) by NVIDIA? offer the option for parallel computations and speed improvements in such high-density processes. With increasing seismic imaging space, the problems associated with multi-GPU techniques need to be addressed. We propose an efficient scheme for multi-GPU programming based on the features of the compute-unified device Architecture(CUDA) using GPU hardware, including concurrent kernel execution, CUDA streams, and peer-to-peer(P2P) communication between the different GPUs. In addition, by adjusting the computing time for imaging during RTM, the data communication times between GPUs become negligible. This means that the overall computation efficiency improves linearly, as the number of GPUs increases. We introduce the multi-GPU scheme by using the acoustic wave propagation and then describe the implementation of RTM in tilted transversely isotropic(TTI) media. Next, we compare the multi-GPU and the unified memory schemes. The results suggest that the proposed multiGPU scheme is superior and, with increasing number of GPUs, the computational efficiency improves linearly.
引文
Cheng,J.,Grossman,M.,and McKercher,T.,2014,Professional CUDA C Programming,Wiley&Sons Inc.,Indianapolis,Indiana,P71,73,74,268,391
Foltinek,D.,Eaton,D.,Mahovsky,J.,Moghaddam,P.,and McGarry,R.,2009,Industrial-scale reverse time migration on GPU hardware:79th Annual International Meeting,SEG,Expanded Abstracts,2789-2793.
Fletcher R.P.,X.Du,P.J.Fowler,2009,Reverse time migration in tilted transversely isotropic media:Geophysics,74,179-187.
Liu,H.W.,Li,B.,Liu,H.,Tong,X.L.,Liu,Q.,Wang,X.W.,and Liu,W.Q.,2012.The issue of prestack reverse time migration and solutions with Graphic Processing Unit implementation,Geophysical Prospecting,60,906-918,doi:10.1111/j.1365-2478.2011.01032.x
Liu,H.W.,Li,B.,Liu,H.,Tong,X.L.and Liu,Q.,2010.The algorithm of high order finite difference pre-stack reverse time migration andGPU implementation.The Chinese Journal of Geophysics,53(7),1725-1733.
Li,B.,Liu,G.F.and Liu,H.,2009.A method of using GPU to accelerate seismic pre-stack time migration:The Chinese Journal of Geophysics,52(1),245-252.
Liu,G.F.,and Li,C.,2016,Practical implementation of prestack Kirchhoff time migration on a general purpose graphic processing unit:Acta Geophysica,64,1051-1063,doi:10.1515/cgeo-2016-0033
Liu,L,R.,Ding,W.,Liu,H.W.,and Liu,H.,2015,3D hybrid-domain full waveform inversion on GPU:Computers and Geosciences,83,27-36.
Liu,G.F.,Meng,X.H.,and Liu,H.,2012,Accelerating finite difference wavefield-continuation depth migration by GPU:Applied Geophysics,9,41-48
Liu,G.F.,Liu,Y.N.,and Meng,X.H.,2013,3D seismic reverse time migration on GPGPU:Computers and Geosciences,59,17-23
Micikevicius,P.,2009,3D Finite difference computation on GPUs using CUDA:Proceedings of 2nd Workshop on General Purpose Processing on Graphics Processing Units,GPGPU-2,Association for Computing Machinery,79-84.
Morton,S.,Cullison,T.,and Micikevicius,P.,2008,Experiences with seismic imaging on GPUs:70th Annual International Conference and exhibition,EAGE,Extended Abstracts,W08.
Nakata,N.,Tsuji,T.and Matsuoka,T.,2011,Acceleration of computation speed for elastic wave simulation using a graphics processing unit:Exploration Geophysics,42,98-104,doi:10.1071/EG10039.
NVIDIA?,2017,CUDA C Programming Guide;http://docs.nvidia.com/cuda/pdf/CUDA_C_Programming_Guide.pdf
Komatitsch,D.,Erlebacher,G.,G?ddeke,D.,and Micheá,D.,2010,High-order finite-element seismic wave propagation modeling with MPIon a large GPUcluster:Journal of Computational Physics,229,7692-7714,doi:10.1016/j.jcp.2010.06.024
Okamoto,T.,Takenaka,H.,Nakamura,T.,and Aoki,T.,2010,Accelerating large-scale simulation of seismic wave propagation by multi-GPUs and threedimensional domain decomposition:Earth Planets and Space,62,939-942
Shi,X.H.,Li,C.,and Wang,S.H.,2011,Computing prestack Kirchhoff time migration on general purpose GPU:Computers&Geosciences,37,1702-1710
Shi,Y and Wang,Y.H.,2016,Reverse time migration of3D vertical seismic profile data:Geophysics,81(1),S31-S38
Weiss,R.M.,and Shragge,J.,2013,solving 3D anisotropic elastic wave equation on parallel GPU devices:Geophysics,78,7-15,doi:10.1190/GEO2012-0063.1
Foltinek,D.,Eaton,D.,Mahovsky,J.,Moghaddam,P.,and McGarry,R.,2009,Industrial-scale reverse time migration on GPU hardware:79th Annual International Meeting,SEG,Expanded Abstracts,2789-2793.
Fletcher R.P.,X.Du,P.J.Fowler,2009,Reverse time migration in tilted transversely isotropic media:Geophysics,74,179-187.
Liu,H.W.,Li,B.,Liu,H.,Tong,X.L.,Liu,Q.,Wang,X.W.,and Liu,W.Q.,2012.The issue of prestack reverse time migration and solutions with Graphic Processing Unit implementation,Geophysical Prospecting,60,906-918,doi:10.1111/j.1365-2478.2011.01032.x
Liu,H.W.,Li,B.,Liu,H.,Tong,X.L.and Liu,Q.,2010.The algorithm of high order finite difference pre-stack reverse time migration andGPU implementation.The Chinese Journal of Geophysics,53(7),1725-1733.
Li,B.,Liu,G.F.and Liu,H.,2009.A method of using GPU to accelerate seismic pre-stack time migration:The Chinese Journal of Geophysics,52(1),245-252.
Liu,G.F.,and Li,C.,2016,Practical implementation of prestack Kirchhoff time migration on a general purpose graphic processing unit:Acta Geophysica,64,1051-1063,doi:10.1515/cgeo-2016-0033
Liu,L,R.,Ding,W.,Liu,H.W.,and Liu,H.,2015,3D hybrid-domain full waveform inversion on GPU:Computers and Geosciences,83,27-36.
Liu,G.F.,Meng,X.H.,and Liu,H.,2012,Accelerating finite difference wavefield-continuation depth migration by GPU:Applied Geophysics,9,41-48
Liu,G.F.,Liu,Y.N.,and Meng,X.H.,2013,3D seismic reverse time migration on GPGPU:Computers and Geosciences,59,17-23
Micikevicius,P.,2009,3D Finite difference computation on GPUs using CUDA:Proceedings of 2nd Workshop on General Purpose Processing on Graphics Processing Units,GPGPU-2,Association for Computing Machinery,79-84.
Morton,S.,Cullison,T.,and Micikevicius,P.,2008,Experiences with seismic imaging on GPUs:70th Annual International Conference and exhibition,EAGE,Extended Abstracts,W08.
Nakata,N.,Tsuji,T.and Matsuoka,T.,2011,Acceleration of computation speed for elastic wave simulation using a graphics processing unit:Exploration Geophysics,42,98-104,doi:10.1071/EG10039.
NVIDIA?,2017,CUDA C Programming Guide;http://docs.nvidia.com/cuda/pdf/CUDA_C_Programming_Guide.pdf
Komatitsch,D.,Erlebacher,G.,G?ddeke,D.,and Micheá,D.,2010,High-order finite-element seismic wave propagation modeling with MPIon a large GPUcluster:Journal of Computational Physics,229,7692-7714,doi:10.1016/j.jcp.2010.06.024
Okamoto,T.,Takenaka,H.,Nakamura,T.,and Aoki,T.,2010,Accelerating large-scale simulation of seismic wave propagation by multi-GPUs and threedimensional domain decomposition:Earth Planets and Space,62,939-942
Shi,X.H.,Li,C.,and Wang,S.H.,2011,Computing prestack Kirchhoff time migration on general purpose GPU:Computers&Geosciences,37,1702-1710
Shi,Y and Wang,Y.H.,2016,Reverse time migration of3D vertical seismic profile data:Geophysics,81(1),S31-S38
Weiss,R.M.,and Shragge,J.,2013,solving 3D anisotropic elastic wave equation on parallel GPU devices:Geophysics,78,7-15,doi:10.1190/GEO2012-0063.1