利用CUDA的剪切波数据三维可视化
|
摘要
提出了一种基于CUDA的并行体绘制技术,将CPU与GPU进行协同工作,将光线投射这一高度并行的过程交由GPU来处理,同时设计了相应的体绘制数据访存优化策略来提高穿并行加速比。实验证明本方法可以将串行绘制速度提升三十倍。通过中国大陆南北带地区的剪切波数据三维可视化实验结果,可以直观看出低温物体的下沉俯冲带的分布。从三维图像可以看出,由于俯冲带的拖曳作用,使得软流圈的物质存在着向上流动的趋势,该高温物质在上升的过程中,使研究区两侧受到热物质的侵蚀而产生拉张作用,在一定程度上导致了该两侧的构造较为活跃,形成了类似断层、裂谷等拉张构造。相对于传统的显示方法,剪切波速度三维可视化更加高效、直观地展示了太平洋俯冲板块对南北带地区和华北克拉通动力学构造的影响,在地球物理学研究中得到良好的应用。
By analyzing the arrival time and the morphology of a shear wave,we can make accurate predictions of underground structures,depth,and morphology.Explanation and elaboration of a shear wave are currently based on qualitative results and two-dimensional display of the images,and not beneficial for analysis of underground structures directly.This paper proposes a parallel volume rendering technology based on CUDA,which makes a CPU and GPU work together.Highly parallel processes such as light sampling are handled by the GPU.Further,we also designed a corresponding visit and access optimizing strategy for rendering volume data:texture register strategies,global register policies,global and shared register strategies,and constant registers strategies.An experiment demonstrates that this model is three times faster than the serial programming with improved visualization results.In this experiment,at the center of the corresponding 3D image,the trends for sinking lowtemperature objects and rising sthenospheric material can be observed.During a trend,two sides generate a tensile function when experiencing erosion from hot material,causing the structure of the two sides to become more active,forming tensile tectonics such as faults and rifts.Using our proposed technology,underground structures and geodynamic information can be explored more effectively.
By analyzing the arrival time and the morphology of a shear wave,we can make accurate predictions of underground structures,depth,and morphology.Explanation and elaboration of a shear wave are currently based on qualitative results and two-dimensional display of the images,and not beneficial for analysis of underground structures directly.This paper proposes a parallel volume rendering technology based on CUDA,which makes a CPU and GPU work together.Highly parallel processes such as light sampling are handled by the GPU.Further,we also designed a corresponding visit and access optimizing strategy for rendering volume data:texture register strategies,global register policies,global and shared register strategies,and constant registers strategies.An experiment demonstrates that this model is three times faster than the serial programming with improved visualization results.In this experiment,at the center of the corresponding 3D image,the trends for sinking lowtemperature objects and rising sthenospheric material can be observed.During a trend,two sides generate a tensile function when experiencing erosion from hot material,causing the structure of the two sides to become more active,forming tensile tectonics such as faults and rifts.Using our proposed technology,underground structures and geodynamic information can be explored more effectively.
引文
[1]Wikipedia.S波[OL].http://zh.wikipedia.org/wiki/S%E6%B3%A2,2003
[2]Bnoit M H,Nyblade A A,Decar J C V.UpperMantle P Wave Speed Variations Beneath Ethiopiaand the Origin of the Afar Hotspot[J].Geology,2006,34(2):329-332
[3]James H,Kostrova T.Use of Lighting in the 3DDisplay of Seismic Data[J].First Break,2005,23(3):53-57
[4]Castanie L,Levy B,Bosquet F.Volume Explorer:Roaming Large Volumes to Couple Visualizationand Data Processing for Oil and Gas Exploration[C].The 16th IEEE Visualization,Minneapolis,USA,2005
[5]Kidd G D.Fundamentals of 3DSeismic Volume Vi-sualization[J].The Leading Edge,1999,18(6):702-709
[6]张学民,刁桂苓,赵英萍,等.华北地区深部地幔剪切波速度结构研究[J].地球物理学报,2006,49(6):1 709-1 719
[7]王椿镛,楼海,吕智勇,等.青藏高原东部地壳上地幔S波速度结构—下地壳流的深部环境[J].中国科学(D辑:地球科学),2008,38(1):22-32
[8]郑勇,葛粲,谢祖军,等.芦山与汶川地震震区地壳上地幔结构及深部孕震环境[J].中国科学(D辑:地球科学),2013,43(6):1 027-1 037
[9]Strengert M,Magallón M,Weiskopf D,et al.Large Volume Visualization of Compressed Time-Dependent Datasets on GPU Clusters[J].ParallelComputing,2005,31(2):205-219
[10]Strengert M,Magallón M,Weiskopf D,et al.Hi-erarchical Visualization and Compression of LargeVolume Datasets Using GPU Clusters[C].Euro-graphics Symposium on Parallel Graphics and Visu-alization(EGPGV),Switzerland,2004
[11]Levoy M.Display of Surfaces from Volume Data[J].IEEE Computer Graphics and Applications,1998,8(3):29-37
[12]NVIDIA Corporation.NVIDIA CUDA Compute U-nified Device Architecture Programming Guide[OL].http://moss.csc.ncsu.edu/~mueller/clus-ter/nvidia/0.8/NVIDIA_CUDA_Programming_Guide_0.8.2.pdf,2008
[2]Bnoit M H,Nyblade A A,Decar J C V.UpperMantle P Wave Speed Variations Beneath Ethiopiaand the Origin of the Afar Hotspot[J].Geology,2006,34(2):329-332
[3]James H,Kostrova T.Use of Lighting in the 3DDisplay of Seismic Data[J].First Break,2005,23(3):53-57
[4]Castanie L,Levy B,Bosquet F.Volume Explorer:Roaming Large Volumes to Couple Visualizationand Data Processing for Oil and Gas Exploration[C].The 16th IEEE Visualization,Minneapolis,USA,2005
[5]Kidd G D.Fundamentals of 3DSeismic Volume Vi-sualization[J].The Leading Edge,1999,18(6):702-709
[6]张学民,刁桂苓,赵英萍,等.华北地区深部地幔剪切波速度结构研究[J].地球物理学报,2006,49(6):1 709-1 719
[7]王椿镛,楼海,吕智勇,等.青藏高原东部地壳上地幔S波速度结构—下地壳流的深部环境[J].中国科学(D辑:地球科学),2008,38(1):22-32
[8]郑勇,葛粲,谢祖军,等.芦山与汶川地震震区地壳上地幔结构及深部孕震环境[J].中国科学(D辑:地球科学),2013,43(6):1 027-1 037
[9]Strengert M,Magallón M,Weiskopf D,et al.Large Volume Visualization of Compressed Time-Dependent Datasets on GPU Clusters[J].ParallelComputing,2005,31(2):205-219
[10]Strengert M,Magallón M,Weiskopf D,et al.Hi-erarchical Visualization and Compression of LargeVolume Datasets Using GPU Clusters[C].Euro-graphics Symposium on Parallel Graphics and Visu-alization(EGPGV),Switzerland,2004
[11]Levoy M.Display of Surfaces from Volume Data[J].IEEE Computer Graphics and Applications,1998,8(3):29-37
[12]NVIDIA Corporation.NVIDIA CUDA Compute U-nified Device Architecture Programming Guide[OL].http://moss.csc.ncsu.edu/~mueller/clus-ter/nvidia/0.8/NVIDIA_CUDA_Programming_Guide_0.8.2.pdf,2008
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