Accelerating the Data Pre-Processing In Beam Migration On A CPU-GPU Hybrid Platform

tengpeng wei,yingqiao wang,haohuan fu,guangwen yang,zhaolin zhu,are osen
DOI: https://doi.org/10.1190/igcbeijing2014-134
2014-01-01
Abstract:PreviousNext No AccessBeijing 2014 International Geophysical Conference & Exposition, Beijing, China, 21-24 April 2014Accelerating the Data Pre-Processing In Beam Migration On A CPU-GPU Hybrid PlatformAuthors: Tengpeng Wei*Yingqiao WangHaohuan FuGuangwen YangZhaolin ZhuAre OsenTengpeng Wei*Tsinghua UniversitySearch for more papers by this author, Yingqiao WangTsinghua UniversitySearch for more papers by this author, Haohuan FuTsinghua UniversitySearch for more papers by this author, Guangwen YangTsinghua UniversitySearch for more papers by this author, Zhaolin ZhuStatoilSearch for more papers by this author, and Are OsenStatoilSearch for more papers by this authorhttps://doi.org/10.1190/IGCBeijing2014-134 SectionsAboutPDF/ePub ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InRedditEmail Abstract Beam migration methods have been shown to keep the same or even better image quality than Kirchhoff migration, with one order of magnitude faster migration speed. However, the data pre-processing step for beam migration, which applies a slant stack operation on all selected local gathers and applies the event picking in the τ-p domain, is still expensive. In this work, we propose a highly-efficient solution of the data preprocessing step in beam migration. Through a fully-pipelined design, we manage to make an efficient utilization of both the CPU and GPU resources in a hybrid platform. Compared with the original version that runs on 12 CPU cores, our proposed hybrid data pre-processing design (12 CPU cores and 2 GPUs) achieves 10 times speedup. The significant performance improvement would further close the gap to an interactive migration engine, and enable more exploration on different event picking strategies. Keywords: beam, migration, Kirchhoff, algorithmPermalink: https://doi.org/10.1190/IGCBeijing2014-134FiguresReferencesRelatedDetails Beijing 2014 International Geophysical Conference & Exposition, Beijing, China, 21-24 April 2014ISSN (online):2159-6832Copyright: 2014 Pages: 1360 publication data© 2014 Published in electronic format with permission by the Society of Exploration Geophysicists and Chinese Petroleum SocietyPublisher:Society of Exploration Geophysicists HistoryPublished Online: 24 Apr 2014 CITATION INFORMATION Tengpeng Wei*, Yingqiao Wang, Haohuan Fu, Guangwen Yang, Zhaolin Zhu, and Are Osen, (2014), "Accelerating the Data Pre-Processing In Beam Migration On A CPU-GPU Hybrid Platform," SEG Global Meeting Abstracts : 524-527. https://doi.org/10.1190/IGCBeijing2014-134 Plain-Language Summary KeywordsbeammigrationKirchhoffalgorithmPDF DownloadLoading ...
What problem does this paper attempt to address?