检索结果-内蒙古大学图书馆

15th International Conference on parallel and distributed Systems, ICPADS '09

作者： Wang, Guibin Tang, Tao Fang, Xudong Ren, Xiaoguang National Laboratory for Parallel and Distributed Processing School of Computer National University of Defense Technology Changsha Hunan China

ISBN: (纸本)9780769539003

Graphic processing Unit (GPU), with many lightweight data-parallel cores, can provide substantial parallel computing power to accelerate several general purpose applications. Both the AMD and NVIDIA corps provide their specific high performance GPUs and software platforms. As the floating-point computing capacity increases continually, the problem of "memory-wall" becomes more serious, especially for array-intensive applications. In this paper, we optimize and implement two SPEC2k benchmarks mgrid and swim on multithreaded GPU using CUDA and Brook+. In order to reduce the pressure on off-hip memory, we make use of data locality in multi-level memory hierarchies and hide long memory access latency via double-buffers. To balance inter-thread parallelism and intra-hread locality, we further tune thread granularity for each kernel and empirically study the best equilibrium point for this problem. Flow control instruction can significantly impact the effective instruction throughput. Oriented to this problem, we introduce a diverge elimination technology to convert condition expression into computing operation. Through all the optimizations, we gain the speedup of 10x-34x to the CPU implementation on the GPUs of AMD and NVIDIA respectively. Finally, we summarize and compares the GPUs from AMD and NVIDIA in hardware and software. © 2009 IEEE.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

Model-directed multi-level parallelization for multi-task programs on the Cell BE

Model-directed multi-level parallelization for multi-task pr...

引用

2009 2nd International Conference on Computer Science and Its Applications, CSA 2009

作者： Wang, Miao Benoit, Nicolas Bodin, François National Laboratory for Parallel and Distributed Processing NUDT Changsha 410073 China University of Rennes 1 Rennes 35000 France

ISBN: (纸本)9781424449460

Multi-core architectures can deliver high processing power if the multiple levels of parallelism they expose are exploited. However, it is non-trivial to orchestrate the computational and memory resources allocation. Furthermore, when dealing with distributed memory architectures, data distribution adds another level of complexity. This paper presents a modeldriven technique for mapping multi-task parallel programs onto multi-core platforms. The resource allocation configurations are expressed within a three-dimensional optimization space which affects task-level and data-level parallelism, and communication. Then, the performance of any valid parallelization scheme is estimated statically, taking into account both computation and communication costs. We prototyped our approach on the Cell BE with an image processing application. Experiments show that our model can correctly predict the overall performance and highlight the most efficient parallelization schemes. © 2009 IEEE.

关键词： Resource allocation

来源：评论

学校读者我要写书评

暂无评论

Computing Must and May Alias to Detect Null Pointer Dereference

Communications in Computer and Information Science

引用

Communications in Computer and Information Science 2009年 17卷 252-261页

作者： Ma, Xiaodong Wang, Ji Dong, Wei National Laboratory for Parallel and Distributed Processing China

ISBN: (纸本)9783540884781

This paper presents a novel algorithm to detect null pointer dereference errors. The algorithm utilizes both of the must and may alias information in a compact way to improve the precision of the detection. Using may alias information obtained by a fast flow- and context- insensitive analysis algorithm, we compute the must alias generated by the assignment statements and the must alias information is also used to improve the precision of the may alias. We can strong update more expressions using the must alias information, which will reduce the false positives of the detection for null pointer dereference. We have implemented our algorithm in the SUIF2 compiler infrastructure and the experiments results are as expected. © Springer-Verlag Berlin Heidelberg 2008.

关键词： Information use

来源：评论

学校读者我要写书评

暂无评论

A Novel network request scheduler for a large scale storage system

A Novel network request scheduler for a large scale storage ...

引用

作者： Qian, Yingjin Barton, Eric Wang, Tom Puntambekar, Nirant Dilger, Andreas National Laboratory for Parallel and Distributed Processing Changsha China Sun Microsystems Inc Lustre Group PH United States

This paper presents a novel Network Request Scheduler (NRS) for a large-scale, LustreTM storage system. It proposes a quantum-based, Object Based Round Robin (OBRR) NRS algorithm that reorders the execution of I/O requests per data object, presenting a workload to backend storage that can be optimized more easily. According to the drawback of static deadlines in large-scale workloads, it proposes a novel two-level deadline setting strategy that not only avoids starvation, but also guarantees that urgent I/O requests are serviced in a specified time period. Via a series of simulation experiments using a Lustre simulator, it demonstrates that I/O performance increases as much as 40% when using the OBRR NRS algorithm, and the two-level deadline setting strategy can avoid starvation and ensure that urgent I/O requests are serviced in the required time. © 2009 Springer-Verlag.

关键词： Scheduling

来源：评论

学校读者我要写书评

暂无评论

Efficient social network approximate analysis on blogosphere based on network structure characteristics

Efficient social network approximate analysis on blogosphere...

引用

3rd Workshop on Social Network Mining and Analysis, SNA-KDD '09

作者： Tang, Jintao Wang, Ting Wang, Ji Wei, Dengping School of Computer National University of Defense Technology ChangSha Hunan 410073 China National Laboratory for Parallel and Distributed Processing ChangSha Hunan 410073 China

ISBN: (纸本)9781605586762

As a typical social media in Web 2.0, blogs have attracted a surge of researches. Unlike the traditional studies, the social networks mined from Internet are very large, which makes a lot of social network analyzing algorithms to be intractable. According to this phenomenon, this paper addresses the novel problem of efficient social networks analyzing on blogs. This paper turns to account the structural characteristics of real large-scale complex networks, and proposes a novel shortest path approximate algorithm to calculate the distance and shortest path between nodes efficiently. The approximate algorithm then is incorporated with social network analysis algorithms and measurements for large-scale social networks analysis. We illustrate the advantages of the approximate analysis through the centrality measurements and community mining algorithms. The experiments demonstrate the effectiveness of the proposed algorithms on blogs, which indicates the necessity of taking account of the structural characteristics of complex networks when optimizing the analysis algorithms on large-scale social networks. Copyright 2009 ACM.

关键词： Blogs

来源：评论

学校读者我要写书评

暂无评论

Quasimonoenergetic Proton Bunch Generation by Dual-Peaked Electrostatic-Field Acceleration in Foils Irradiated by an Intense Linearly Polarized Laser

引用

Physical Review Letters 2010年第6期105卷 065003-065003页

作者： H. B. Zhuo (卓红斌) Z. L. Chen (陈正林) W. Yu (余玮) Z. M. Sheng (盛政明) M. Y. Yu (郁明阳) Z. Jin (金展) R. Kodama (兒玉了祐) Graduate School of Engineering Osaka University 2-1 Yamada-oka Suita Osaka 565-0871 Japan National Laboratory for Parallel and Distributed Processing School of Computer Science National University of Defense Technology Changsha 410073 People’s Republic of China Shanghai Institute of Optics and Fine Mechanics Chinese Academy of Sciences Shanghai 201800 People’s Republic of China Department of Physics Shanghai Jiao Tong University Shanghai 200240 People’s Republic of China Beijing National Laboratory for Condensed Matter Physics Institute of Physics CAS Beijing 100080 People’s Republic of China Institute for Fusion Theory and Simulation Zhejiang University Hangzhou 310027 People’s Republic of China Institute for Theoretical Physics I Ruhr University Bochum D-44780 Germany Institute of Laser Engineering Osaka University 2-6 Yamada-oka Suita Osaka 565-0871 Japan

It is found that stable proton acceleration from a thin foil irradiated by a linearly polarized ultraintense laser can be realized for appropriate foil thickness and laser intensity. A dual-peaked electrostatic field, originating from the oscillating and nonoscillating components of the laser ponderomotive force, is formed around the foil surfaces. This field combines radiation-pressure acceleration and target normal sheath acceleration to produce a single quasimonoenergetic ion bunch. A criterion for this mechanism to be operative is obtained and verified by two-dimensional particle-in-cell simulation. At a laser intensity of ∼5.5×1022 W/cm2, quasimonoenergetic GeV proton bunches are obtained with ∼100 MeV energy spread, less than 4° spatial divergence, and ∼50% energy conversion efficiency from the laser.

关键词： Germanium compounds

来源：评论

学校读者我要写书评

暂无评论

Mobility of Internet-Based Virtual Computing Environment

Mobility of Internet-Based Virtual Computing Environment

引用

International Conference on parallel and distributed Systems (ICPADS)

作者： Siqi Shen Ji Wang Rui Shen Shengdong Zhang Pei Fan National Laboratory for Parallel and Distributed Processing Changsha China

The Internet-based virtual computing environment (iVCE) provides on-demand aggregation and autonomic collaboration mechanisms to facilitate the utilization of autonomous and dynamic Internet resources. Load balancing and fault tolerance are important issues when scheduling those transient resources. In this paper, we propose a mobility mechanism for the migration of various roles of agents in the iVCE platform. The mobility mechanism involves two parts of the iVCE platform: role container layer and event service layer. At the role container layer, a novel approach is proposed to handle the code and data mobility issue. At the event service layer, an efficient routing reconfiguration protocol is proposed based on a publish/subscribe system over DHTs to facilitate task migrations. Certain conditions must be satisfied before the migration of an agent to ensure the correctness of the whole process. Experiments are conducted to evaluate the performance of the mobility mechanism, and the experimental results show that it is suitable for implementing load balancing and fault tolerance in the iVCE.

关键词： Internet Load management Fault tolerance Containers distributed computing Computer languages Concurrent computing International collaboration Routing protocols Middleware

来源：评论

学校读者我要写书评

暂无评论

A framework for constructing peer-to-peer overlay networks in java 09

A framework for constructing peer-to-peer overlay networks i...

引用

7th International Conference on Principles and Practice of Programming in Java, PPPJ 2009

作者： Rui, Shen Ji, Wang Shengdong, Zhang Siqi, Shen Pei, Fan National Laboratory for Parallel and Distributed Processing Changsha 410073 China School of Computing University of Leeds Leeds LS2 9JT United Kingdom

ISBN: (纸本)9781605585987

Peer-to-peer emerges as a better way for building applications on the Internet that require high scalability and availability. Peer-to-peer systems are usually organized into structured overlay networks, which provide key-based routing capabilities to eliminate flooding in unstructured ones. Many overlay network protocols have been proposed to organize peers into various topologies with emphasis on different networking properties. However, applications are often stuck to a specific peer-to-peer overlay network implementation, because different overlay implementations usually provide very different interfaces and messaging mechanisms. In this paper, we present a framework for constructing peer-to-peer overlay networks in Java. First, networking is abstracted by the interfaces that use URIs to uniformly address peers on different underlying or overlay networks. Then, asynchronous and synchronous messaging support is built upon these interfaces. Finally, overlay networking interfaces are sketched to handle specific issues in overlay networks. We have constructed several overlay networks in this framework, and built peer-to-peer applications which are independent of overlay implementations. Copyright 2009 ACM.

关键词： Overlay networks

来源：评论

学校读者我要写书评

暂无评论

A peer-to-peer IO buffering service based on RAM-grid

引用

International Journal of Autonomous and Adaptive Communications Systems 2009年第4期2卷 382-396页

作者： Zhang, Yiming Chu, Rui Li, Dongsheng National Laboratory for Parallel and Distributed Processing National University of Defense Technology Changsha 410073 China

The performance of IO-intensive applications is determined by the hit ratio of local disk cache and the IO latency of missed disk accesses. To improve the IO performance, in this paper we propose PIB, a peer-to-peer IO buffering service based on RAM-grid (a new grid system aiming at memory resource sharing in WAN). The PIB service acts as a two-level disk cache, which buffers obsolete blocks in idle nodes on the internet for IO-intensive applications. It reduces the IO latency of missed disk accesses by means of the speed advantage of network over disks, and improves the hit ratio of local cache based on accurate identification of IO patterns. The effectiveness of our proposals is demonstrated through the trace driven simulation studies. Copyright © 2009 Inderscience Enterprises Ltd.

关键词： Random access storage

来源：评论

学校读者我要写书评

暂无评论

An irregular cutting based distributed mapping algorithm for DVE systems

An irregular cutting based distributed mapping algorithm for...

引用

2009 2nd Asia-Pacific Conference on Computational Intelligence and Industrial Applications, PACIIA 2009

作者： Zhang, Wei Zhou, Hangjun Peng, Yuxing Li, Sikun Xi, Min National Laboratory for Parallel and Distributed Processing School of Computer Science National University of Defense Technology Changsha 410073 China Department of Computer Science Xi'an Jiaotong University Xi'an 710049 China

ISBN: (纸本)9781424446070

distributed Virtual Environment (DVE) systems have become more and more important both in academic communities and the industries. To guarantee the load constrain, the physical world integrity and the virtual world integrity, the virtual environment should be divided into several parts and mapped to multiple servers, which are located in geographically different places. Many centralized algorithms based on regular cutting method are proposed to solve this problem. However, the dynamic property of clients and distributed property of the system make the existing algorithms unable to undertake it. In this paper, we propose a metric to measure the cutting quality and design a distributed mapping algorithm based on an irregular cutting. Simulation results show that our algorithm greatly decreases the costs of DVE systems for migration of clients and effectively improves the satisfaction degree of users. ©2009 IEEE.

关键词： Virtual reality

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：