基于CUDA的Kirchhoff叠前时间偏移算法设计与实现
|
摘要
Kirchhoff叠前时间偏移是地震数据处理中最耗时的常用模块之一。为加快计算和显示速度,针对CU-DA平台多处理器流水线特性,对传统Kirchhoff叠前时间偏移算法在CUDA平台上进行了重新设计,包括基于CUDA的Kirchhoff叠前时间偏移算法、基于CUDA的纵波波动方程算法和GPU与CPU间的通信算法三个子算法。所有算法在NVIDIAGeForce8800 GT系统上编译实现,通过对比相同数据在Intel Core2Due CPU2.0 GHz的地震偏移,综合分析和实验结果表明,基于CUDA的Kirchhoff叠前时间偏移算法的计算速度最高可得到较传统CPU算法6倍的加速比,从而为快速处理地震偏移提供良好的支持。
Kirchhoff prestack time migration is the most time-consuming part of seismic data process method.In order to accelerating seismic data computing and displaying,the paper designed prestack Kirchhoff time migration algorithm on CUDA platform considering multi-processor line features,including CUDA-based Kirchhoff prestack time migration algorithm,CUDA-based wave equation algorithm,GPU and CPU communication algorithm.Conducted extensive experiments in a PC with Intel Core2Due CPU 2.0 GHz and NVIDIA GeForce 8800 GT graphic card.Presented a comprehensive performance study to prove the efficiency of the proposed algorithms,CUDA-based Kirchhoff prestack time migration algorithm 6 times faster than available traditional CPU algorithm.The results obtained here indicate that GPU can potentially be a very useful platform for processing large-scale seismic data.
Kirchhoff prestack time migration is the most time-consuming part of seismic data process method.In order to accelerating seismic data computing and displaying,the paper designed prestack Kirchhoff time migration algorithm on CUDA platform considering multi-processor line features,including CUDA-based Kirchhoff prestack time migration algorithm,CUDA-based wave equation algorithm,GPU and CPU communication algorithm.Conducted extensive experiments in a PC with Intel Core2Due CPU 2.0 GHz and NVIDIA GeForce 8800 GT graphic card.Presented a comprehensive performance study to prove the efficiency of the proposed algorithms,CUDA-based Kirchhoff prestack time migration algorithm 6 times faster than available traditional CPU algorithm.The results obtained here indicate that GPU can potentially be a very useful platform for processing large-scale seismic data.
引文
[1]GRAY S H,ETGEN J,DELLINGER J,et al.Seismic migrationproblems and solutions[J].Geophysics Wisconsin then TulsaSociety of Exploration,2001,66(5):1622-1640.
[2]HE Chuan,LU Mi,SUN Chuan-wen.Accelerating seismic migrationusing FPGA-based coprocessor platform[C]//Proc of the 12th AnnualIEEE Symposium on Field-Programmable Custom Computing Ma-chines.Washington DC:IEEE Computer Society,2004:207-216.
[3]PANETTA J,FILHO P,FILHO C,et al.Computational characteris-tics of production seismic migration and its performance on novel pro-cessor architectures[C]//Proc of the 19th International Symposiumon Computer Architecture and High Performance Computing.2007:11-18.
[4]MIZUKAMI Y,TADAMURA K.Optical flow computation on com-pute unified device architecture[C]//Proc of the 14th IEEE Interna-tional Conference on Image Analysis and Processing.2007:179-184.
[5]Al-KISWANY S,GHARAIBEH A,SANTOS-NTO E,et al.Store-GPU:exploiting graphics processing units to accelerate distributedstorage systems[C]//Proc of HPDC’08.New York:ACM Press,2008:165-174.
[6]LESSIG C.An implementation of the MRRR algorithm on a data pa-rallel coprocessor[R].Toronto:University of Toronto,2008.
[7]TOLKE J.Implementation of a lattice boltzmann kernel using thecompute unified device architecture developed by NVIDIA[C]//Procof Conference on Computing Frontiers.2008.
[8]CUDA programming guide 1.1[K].[S.l.]:NVIDIA,2007.
[9]BEVC D.Imaging complex structure with semirecursive Kirchhoff mi-gration[J].Geophysics,1997,62(2):577-588.
[10]YE Yue-ming,LI Zhen-chun,HAN Wen-gong,et al.Beamlet pres-tack depth migration and illumination:a test based on the Marmousimodel[J].Applied Geophyisics,2006,3(4):203-209.
[11]吴时国,王秀铃,季玉新,等.3DMove构造裂缝预测技术在古潜山的应用研究[J].中国科学D辑:地球科学,2004,34(9):795-806.
[12]陈志德,刘宽,李成斌.三维叠前深度偏移速度分析及蒙卡洛自动层速度拾取[J].地球物理学报,2002,45(2):246-254.
[13]李景叶,陈小宏,芮振华.基于匹配滤波的多次波压制方法研究[J].地球物理学进展,2007,22(1):200-206.
[14]朱良峰,潘信,吴信才,等.地质断层三维可视化模型的构建方法与实现技术[J].软件学报,2008,19(8):2004-2017.
[2]HE Chuan,LU Mi,SUN Chuan-wen.Accelerating seismic migrationusing FPGA-based coprocessor platform[C]//Proc of the 12th AnnualIEEE Symposium on Field-Programmable Custom Computing Ma-chines.Washington DC:IEEE Computer Society,2004:207-216.
[3]PANETTA J,FILHO P,FILHO C,et al.Computational characteris-tics of production seismic migration and its performance on novel pro-cessor architectures[C]//Proc of the 19th International Symposiumon Computer Architecture and High Performance Computing.2007:11-18.
[4]MIZUKAMI Y,TADAMURA K.Optical flow computation on com-pute unified device architecture[C]//Proc of the 14th IEEE Interna-tional Conference on Image Analysis and Processing.2007:179-184.
[5]Al-KISWANY S,GHARAIBEH A,SANTOS-NTO E,et al.Store-GPU:exploiting graphics processing units to accelerate distributedstorage systems[C]//Proc of HPDC’08.New York:ACM Press,2008:165-174.
[6]LESSIG C.An implementation of the MRRR algorithm on a data pa-rallel coprocessor[R].Toronto:University of Toronto,2008.
[7]TOLKE J.Implementation of a lattice boltzmann kernel using thecompute unified device architecture developed by NVIDIA[C]//Procof Conference on Computing Frontiers.2008.
[8]CUDA programming guide 1.1[K].[S.l.]:NVIDIA,2007.
[9]BEVC D.Imaging complex structure with semirecursive Kirchhoff mi-gration[J].Geophysics,1997,62(2):577-588.
[10]YE Yue-ming,LI Zhen-chun,HAN Wen-gong,et al.Beamlet pres-tack depth migration and illumination:a test based on the Marmousimodel[J].Applied Geophyisics,2006,3(4):203-209.
[11]吴时国,王秀铃,季玉新,等.3DMove构造裂缝预测技术在古潜山的应用研究[J].中国科学D辑:地球科学,2004,34(9):795-806.
[12]陈志德,刘宽,李成斌.三维叠前深度偏移速度分析及蒙卡洛自动层速度拾取[J].地球物理学报,2002,45(2):246-254.
[13]李景叶,陈小宏,芮振华.基于匹配滤波的多次波压制方法研究[J].地球物理学进展,2007,22(1):200-206.
[14]朱良峰,潘信,吴信才,等.地质断层三维可视化模型的构建方法与实现技术[J].软件学报,2008,19(8):2004-2017.
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