高性能计算的发展现状及趋势
|
摘要
地震勘探是高性能计算最重要的应用领域之一。本文通过跟踪国内外高性能计算机的发展动态,分析地球物理勘探领域对高性能计算的需求和应用现状,展望高性能计算的发展趋势,得出了以下结论和认识:高性能计算机今后发展趋势是向多核方向发展;②目前地震勘探对高性能计算的直接需求是叠前深度偏移、叠前反演和全波形反演,以及单点高密度勘探技术;③GPU技术具有广阔的应用前景,尤其是新建的CUDA架构完善了软件开发环境;④基于FPGAs(Field Programmable Gate Arrays,现场可编程门阵列)的可重构技术,既有硬件的计算性能,又有软件的灵活性,采用此项技术可以大大提高计算速度;⑤受功耗、传统集成电路技术等制约,单CPU性能提高有很大的局限性,开发新材料、完善计算机封装结构成为提高计算性能的新途径,光电二极管(APD)硅光子器件、内存+CPU的MCP封装结构,发展前景看好;⑥云计算(Cloud Computing)是在分布计算、并行计算、网格计算等基础上发展起来的一种全新的数据密集型超级计算模式,该技术将在海量数据的高性能计算方面有很好的应用前景;⑦计算性能的提高依赖于软硬件一体化整体发展,目前硬件技术发展日新月异,但软件开发及应用相对滞后,大力发展软件产业是解决以上问题的必由之路。
Seismic exploration is one of the most important application areas for high performance computing.By tracking the development trends of high performance computers at home and aboard,analyzing the requirements and application status of geophysical exploration to high performance computing,and forecasting the development trends of the high performance computing,the following conclusions and understandings were obtained in this paper:(1):Multi-cores will be the future development trends for high performance computing,(2):High performance computing is needed in pre-stack depth migration,pre-stack inversion and full waveform inversion,and single point high density exploration in seismic exploration,(3):GPU technology has wide application prospects,especially the newly built CUDA framework which improves the software development environment,(4):The Reconfigurable Technology which is based on FPGAs(Field Programmable Gate Arrays)not only has hardware's calculation performance,but also has software flexibility,application of this technology could greatly raise the computing speed,(5):Constrained by power consumption and traditional Integrated Circuit Technology,there is significant limitation for single CPU performance improvements,developing new materials and improving computer's package structure become the new ways for raising computer's performance,there should be good future for some developments,such as Avalanche Photoelectric Diode(APD)Silicon photonic devices,memory+CPU MCP package structure,(6):Cloud Computing is a totally new mode of data-intensive supercomputing mode which is based on distribution computing,parallel computing and grid computing,the technology will have a good application prospect in high performance computing for vast amount of data.(7):The raising of the computing performance depends on integrated development of hardware and software,compared with fast development in hardware,the development and application of software were left behind,as a result greatly developing the software industry is the only way to solve the above problems.
Seismic exploration is one of the most important application areas for high performance computing.By tracking the development trends of high performance computers at home and aboard,analyzing the requirements and application status of geophysical exploration to high performance computing,and forecasting the development trends of the high performance computing,the following conclusions and understandings were obtained in this paper:(1):Multi-cores will be the future development trends for high performance computing,(2):High performance computing is needed in pre-stack depth migration,pre-stack inversion and full waveform inversion,and single point high density exploration in seismic exploration,(3):GPU technology has wide application prospects,especially the newly built CUDA framework which improves the software development environment,(4):The Reconfigurable Technology which is based on FPGAs(Field Programmable Gate Arrays)not only has hardware's calculation performance,but also has software flexibility,application of this technology could greatly raise the computing speed,(5):Constrained by power consumption and traditional Integrated Circuit Technology,there is significant limitation for single CPU performance improvements,developing new materials and improving computer's package structure become the new ways for raising computer's performance,there should be good future for some developments,such as Avalanche Photoelectric Diode(APD)Silicon photonic devices,memory+CPU MCP package structure,(6):Cloud Computing is a totally new mode of data-intensive supercomputing mode which is based on distribution computing,parallel computing and grid computing,the technology will have a good application prospect in high performance computing for vast amount of data.(7):The raising of the computing performance depends on integrated development of hardware and software,compared with fast development in hardware,the development and application of software were left behind,as a result greatly developing the software industry is the only way to solve the above problems.
引文
[1]张军华,仝兆岐.地震资料处理中的并行计算机技术.物探化探计算技术,2002,24(1):31～36
[2]张军华,雷凌,仝兆岐.PC Cluster技术的国内外现状与发展趋势.石油物探,2003,42(4):557～561
[3]TOP10Systems.www.top500.org
[4]高性能计算机发展面临三大挑战.http:∥finance.si-na.com.cn/roll/20100317/08233248407.sht ml
[5]超级“大算盘”一天顶160年——国产千万亿次超级计算机亮相.http:∥news.sina.com.cn/o/2009-10-30/075816525368s.sht ml
[6]http:∥www.ronglian.com/solution5.ht ml
[7]Paul Matheny,Robert Sambell,Said Mahrooqi et al.Evolution of the land seismic super crew.SEG Tech-nical Program Expanded Abstracts,2009,28:81～85
[8]William J Camp,Philippe Thierry.Trends for high-performance scientific computing.The Leading Edge,2010,29(1):44～47
[9]赵改善.我们需要多大和多快的计算机.勘探地球物理进展,2004,27(1):22～28
[10]赵改善,孔祥宁,王于静等.64位集群计算平台波动方程叠前深度偏移的性能优化.勘探地球物理进展,2005,28(1):57～64
[11]John W C Sherwood,Kevin Sherwood et al.3D beam prestack depth migration with examples fromaround the word.The Leading Edge,2009,28(9):1120～1127
[12]John Etgen,Samuel H Gray,Zhang Yu.An overview of depthimagingin exploration geophysics.Geophys-ics,2009,WCA5～WCA17
[13]张美根,王妙月,李小凡等.时间域全波形各向异性弹性参数反演.地球物理学报,2003,46(1):94～100
[14]孙银行,乐友喜.各向异性弹性参数的广义非线性反演方法.勘探地球物理进展,2007,30(3):179～184
[15]Denes Vigh,William E Starr,Jerry Kapoor.Develo-ping Earth models with full waveforminversion.The Leading Edge,2009,28(4):432～435
[16]http:∥www.westerngeco.com/services/dp/omega/depth/tomoportfolio/f wi.aspx
[17]钟联波.GPU与CPU的比较分析.技术与市场,2009,16(9):13～14
[18]郭境峰,蔡伟涛.新一代高能运算技术——CUDA简介.现代科技,2009,8(6):58～62
[19]Benjamin J Kadlec,Geoffrey A Dorn.Levering graph-ics processing units(GPUs)for real-time seismic in-terpretatiion.The Leading Edge,2010,29(1):60～66
[20]张兵,赵改善,黄骏等.地震叠前深度偏移在CUDA平台上的实现.勘探地球物理进展,2008,31(6):427～432
[21]赵改善.地球物理高性能计算的新选择:GPU计算技术.勘探地球物理进展,2007,30(5):399～404
[22]Owen J D,Luebke D,Govindaraju Net al.Asurvy of general-purpose computation on graphics hardware.Computer Graphics Forum,2007,26(1):80～113
[23]徐新海,林宇斐,易伟.CPU-GPGPU异构体系结构相关技术综述.计算机工程与科学,2009,31,(A1):24～26
[24]张朝晖,刘俊起,徐勤建.GPU并行计算技术分析与应用.信息技术,2009,11:86～89
[25]Li Bo,Tong Xiao-long,Li You-ming.A method of ac-celerating seismic Pre-stack time migration by GPU.SEG Technical Program Expanded Abstracts,2009,28:3055～3058
[26]赵改善.可重构计算技术及其在地球物理中的应用前景.勘探地球物理进展,2007,30(4):309～316
[27]He Chuan,Lu Mi,Sun Chuan-wen.Accelerating seismic migration using FPGA-based coprocessor platform.∥Proceedings of the12th Annual IEEE Symposium on Field-Programmable Custom Compu-ting Machines,2004,207～216
[28]He Chuan,Sun Chuan-wen,Lu Mi.Prestack Kirchh-off time migration on high performance reconfigurable computing platform.SEG Technical Program Ex-panded Abstracts,2005,24:1902～1906
[29]Robert G Clapp,Haohuan Fu.Selecting the right har dware for reverse time migration.The Leading Edge,2010,29(1):48～58
[30]Fu Hao-huan,Robert G Clapp.Accerating3D Convo-lution using Streaming Architecture on FPGAs.SEG Technical Program Expanded Abstracts,2009,28:3035～3039
[31]日本富士通公司开发出迄今运算速度最快CPU.ht-tp:∥hz.yesky.com/300/8858800.sht ml
[32]基于排队论的单核处理器和多核处理器性能的简单量化评测.http:∥wenku.baidu.com/view/ce1151d380eb6294dd886c06.ht ml
[33]Vaquero L M,Rodero M L,Caceres J et al.A break in the clouds:Toward a cloud definition.ACMSIGCOMM Computer Communication Review,2009,39(1):50～55
[34]我国高性能计算产业,硬件巨人,软件矮子.http:∥hi.baidu.com/daditel/blog/item/
[2]张军华,雷凌,仝兆岐.PC Cluster技术的国内外现状与发展趋势.石油物探,2003,42(4):557～561
[3]TOP10Systems.www.top500.org
[4]高性能计算机发展面临三大挑战.http:∥finance.si-na.com.cn/roll/20100317/08233248407.sht ml
[5]超级“大算盘”一天顶160年——国产千万亿次超级计算机亮相.http:∥news.sina.com.cn/o/2009-10-30/075816525368s.sht ml
[6]http:∥www.ronglian.com/solution5.ht ml
[7]Paul Matheny,Robert Sambell,Said Mahrooqi et al.Evolution of the land seismic super crew.SEG Tech-nical Program Expanded Abstracts,2009,28:81～85
[8]William J Camp,Philippe Thierry.Trends for high-performance scientific computing.The Leading Edge,2010,29(1):44～47
[9]赵改善.我们需要多大和多快的计算机.勘探地球物理进展,2004,27(1):22～28
[10]赵改善,孔祥宁,王于静等.64位集群计算平台波动方程叠前深度偏移的性能优化.勘探地球物理进展,2005,28(1):57～64
[11]John W C Sherwood,Kevin Sherwood et al.3D beam prestack depth migration with examples fromaround the word.The Leading Edge,2009,28(9):1120～1127
[12]John Etgen,Samuel H Gray,Zhang Yu.An overview of depthimagingin exploration geophysics.Geophys-ics,2009,WCA5～WCA17
[13]张美根,王妙月,李小凡等.时间域全波形各向异性弹性参数反演.地球物理学报,2003,46(1):94～100
[14]孙银行,乐友喜.各向异性弹性参数的广义非线性反演方法.勘探地球物理进展,2007,30(3):179～184
[15]Denes Vigh,William E Starr,Jerry Kapoor.Develo-ping Earth models with full waveforminversion.The Leading Edge,2009,28(4):432～435
[16]http:∥www.westerngeco.com/services/dp/omega/depth/tomoportfolio/f wi.aspx
[17]钟联波.GPU与CPU的比较分析.技术与市场,2009,16(9):13～14
[18]郭境峰,蔡伟涛.新一代高能运算技术——CUDA简介.现代科技,2009,8(6):58～62
[19]Benjamin J Kadlec,Geoffrey A Dorn.Levering graph-ics processing units(GPUs)for real-time seismic in-terpretatiion.The Leading Edge,2010,29(1):60～66
[20]张兵,赵改善,黄骏等.地震叠前深度偏移在CUDA平台上的实现.勘探地球物理进展,2008,31(6):427～432
[21]赵改善.地球物理高性能计算的新选择:GPU计算技术.勘探地球物理进展,2007,30(5):399～404
[22]Owen J D,Luebke D,Govindaraju Net al.Asurvy of general-purpose computation on graphics hardware.Computer Graphics Forum,2007,26(1):80～113
[23]徐新海,林宇斐,易伟.CPU-GPGPU异构体系结构相关技术综述.计算机工程与科学,2009,31,(A1):24～26
[24]张朝晖,刘俊起,徐勤建.GPU并行计算技术分析与应用.信息技术,2009,11:86～89
[25]Li Bo,Tong Xiao-long,Li You-ming.A method of ac-celerating seismic Pre-stack time migration by GPU.SEG Technical Program Expanded Abstracts,2009,28:3055～3058
[26]赵改善.可重构计算技术及其在地球物理中的应用前景.勘探地球物理进展,2007,30(4):309～316
[27]He Chuan,Lu Mi,Sun Chuan-wen.Accelerating seismic migration using FPGA-based coprocessor platform.∥Proceedings of the12th Annual IEEE Symposium on Field-Programmable Custom Compu-ting Machines,2004,207～216
[28]He Chuan,Sun Chuan-wen,Lu Mi.Prestack Kirchh-off time migration on high performance reconfigurable computing platform.SEG Technical Program Ex-panded Abstracts,2005,24:1902～1906
[29]Robert G Clapp,Haohuan Fu.Selecting the right har dware for reverse time migration.The Leading Edge,2010,29(1):48～58
[30]Fu Hao-huan,Robert G Clapp.Accerating3D Convo-lution using Streaming Architecture on FPGAs.SEG Technical Program Expanded Abstracts,2009,28:3035～3039
[31]日本富士通公司开发出迄今运算速度最快CPU.ht-tp:∥hz.yesky.com/300/8858800.sht ml
[32]基于排队论的单核处理器和多核处理器性能的简单量化评测.http:∥wenku.baidu.com/view/ce1151d380eb6294dd886c06.ht ml
[33]Vaquero L M,Rodero M L,Caceres J et al.A break in the clouds:Toward a cloud definition.ACMSIGCOMM Computer Communication Review,2009,39(1):50～55
[34]我国高性能计算产业,硬件巨人,软件矮子.http:∥hi.baidu.com/daditel/blog/item/
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