检索结果-内蒙古大学图书馆

GENERATING LOCAL ADDRESSES AND COMMUNICATION SETS FOR data-parallel PROGRAMS

JOURNAL OF parallel AND DISTRIBUTED COMPUTING 1995年第1期26卷 72-84页

作者： CHATTERJEE, S GILBERT, JR LONG, FJE SCHREIBER, R TENG, SH SYST ENGN ASSOCIATES SAN DIEGOCA 92109 MIT DEPT MATHCAMBRIDGEMA 02139 XEROX CORP PALO ALTO RES CTRCOMP SCI LABPALO ALTOCA 94304

Generating local addresses and communication sets is an important issue in distributed-memory implementations of data-parallel languages such as High Performance Fortran. We demonstrate a storage scheme for an array A affinely aligned to a template that is distributed across p processors with a cyclic(k) distribution that does not waste any storage, and show that, under this storage scheme, the local memory access sequence of any processor for a computation involving the regular section A(l:h:s) is characterized by a finite state machine of at most k states. We present fast algorithms for computing the essential information about these state machines, and we extend the framework to handle multidimensional arrays. We also show how to generate communication sets using the state machine approach. Performance results show that this solution requires very little runtime overhead and acceptable preprocessing time. (C) 1995 Academic Press, Inc.

关键词： Communication Research data parallel arrays SETTING finite state machines state machines PROCESSOR Local Memory distributed memory parallel languages jigs(templates) Storage

来源：评论

学校读者我要写书评

暂无评论

Understanding Scalability and Fine-Grain parallelism of Synchronous data parallel Training

Understanding Scalability and Fine-Grain Parallelism of Sync...

引用

IEEE/ACM Workshop on Machine Learning in High Performance Computing Environments (MLHPC)

作者： Jiali Li Bogdan Nicolae Justin Wozniak George Bosilca The University of Tennessee Knoxville USA Argonne National Laboratory USA

ISBN: (纸本)9781728159867

In the age of big data, deep learning has emerged as a powerful tool to extract insight and exploit its value, both in industry and scientific applications. With increasing complexity of learning models and amounts of training data, data-parallel approaches based on frequent all-reduce synchronization steps are increasingly popular. Despite the fact that high-performance computing (HPC) technologies have been designed to address such patterns efficiently, the behavior of data-parallel approaches on HPC platforms is not well understood. To address this issue, in this paper we study the behavior of Horovod, a popular data-parallel approach that relies on MPI, on Theta, a pre-Exascale machine at Argonne National Laboratory. Using two representative applications, we explore two aspects: (1) how performance and scalability is affected by important parameters such as number of nodes, number of workers, threads per node, batch size; (2) how computational phases are interleaved withall-reduce communication phases at fine granularity and what consequences this interleaving has in terms of potential bottlenecks. Our findings show that pipelining of back-propagation, gradient reduction and weight updates mitigate the effects of stragglers during all-reduce only partially. Furthermore, there can be significant delays between weights update, which can be leveraged to mask the overhead of additional background operations that are coupled with the training.

关键词： Deep learning data-parallel training Behavior analysis behavioral research data parallel Scalability High Performance Computing Fine Grain Size ARGONNE NATIONAL LABORATORY Pipeline processing microparticles Training message passing mannose phosphate isomerase remote procedure calls

来源：评论

学校读者我要写书评

暂无评论

Ship-iceberg discrimination from Sentinel-1 synthetic aperture radar data using parallel convolutional neural network

引用

CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE 2021年第17期33卷 e6297-e6297页

作者： Song, Lan Peters, Dennis K. Huang, Weimin Power, Desmond East China Jiaotong Univ Sch Informat Engn Nanchang 330013 Jiangxi Peoples R China Mem Univ Fac Engn & Appl Sci St John NF A1B 3X5 Canada Wuhan Univ Sch Comp Sci Wuhan Peoples R China C CORE St John NF Canada

Ships and icebergs are similar in size and intensity in SAR images, so it is difficult to distinguish them in remote sensing images. Deep learning is a technique based on neural networks, which has played an important role in image information processing. In order to address the challenge of ship and iceberg classification, we present a convolutional neural network (CNN) based classification method for iceberg and ship discrimination from Sentinel-1 SAR images with different polarizations and incidence angles. The method is based on the fixed constant false alarm rate (CFAR) detector and the CNN model has three input channels, then the model was trained using parallel algorithm. The CNN is trained using 1443 images and tested using 161 images. The CNN model is also compared with support vector machine (SVM) and k nearest neighbors (kNN) using the same dataset. Comparison shows the CNN-based method performs the best, and it achieved a validation accuracy of 96%.

关键词： convolution neural network data parallel Sentinel-1 radar images

来源：评论

学校读者我要写书评

暂无评论

A Survey on parallel Computing and its Applications in data-parallel Problems Using GPU Architectures

引用

Communications in Computational Physics 2014年第2期15卷 285-329页

作者： Cristobal A.Navarro Nancy Hitschfeld-Kahler Luis Mateu Department of Computer Science(DCC) Universidad de ChileSantiagoChile Centro de Estudios Cient´ıficos(CECS) ValdiviaChile.

parallel computing has become an important subject in the field of computer science and has proven to be critical when researching high performance *** evolution of computer architectures(multi-core and many-core)towards a higher number of cores can only confirm that parallelism is the method of choice for speeding up an *** the last decade,the graphics processing unit,or GPU,has gained an important place in the field of high performance computing(HPC)because of its low cost and massive parallel processing ***-computing has become,for the first time,available to anyone at the price of a desktop *** this paper,we survey the concept of parallel computing and especially GPU *** efficient parallel algorithms for the GPU is not a trivial task,there are several technical restrictions that must be satisfied in order to achieve the expected *** of these limitations are consequences of the underlying architecture of the GPU and the theoretical models behind *** goal is to present a set of theoretical and technical concepts that are often required to understand the GPU and its massive parallelism *** particular,we show how this new technology can help the field of computational physics,especially when the problem is *** present four examples of computational physics problems;n-body,collision detection,Potts model and cellular automata *** examples well represent the kind of problems that are suitable for GPU *** understanding the GPU architecture and its massive parallelism programming model,one can overcome many of the technical limitations found along the way,design better GPU-based algorithms for computational physics problems and achieve speedups that can reach up to two orders of magnitude when compared to sequential implementations.

关键词： GPU computing parallel computing computing models algorithms data parallel massive parallelism Potts model Ising Model collision detection n-body Cellular Automata.

来源：评论

学校读者我要写书评

暂无评论

data parallel algorithm on geometric correction of remote-sensing satellite images

引用

Jisuanji Yanjiu yu Fazhan/Computer Research and Development 2004年第7期41卷 1200-1206页

作者： Zhang, Fa-Cun Wang, Zhong Zhao, Xiao-Hong Shen, Xu-Bang Xi'an Microelectron. Tech. Inst. Xi'an 710054 China

A real-time algorithm is discussed on geometric correction of a remote-sensing image on satellite. Nowadays, a remote-sensing image is more than 10 thousands of rows and columns larger and the traditional methods of sequential computing on a single processor can not meet the needs of real time processing on satellite. Proposed in the paper is a data parallel algorithm on one dimensional PE array of SIMD computer based on the shift of PE array, and under the parameters of NASA LANDSAT-1 the algorithm is discussed in detail, and the way of implementation is given. By discussing on complexity and speed-up of the algorithm, it is clear that this algorithm is N times higher in efficiency than the sequential algorithm.

关键词： data parallel Geometric correction PE array shift Remote-sensing image

来源：评论

学校读者我要写书评

暂无评论

Efficient data-parallel Algorithm for Elevation Color Generation in Terrain Rendering

Efficient Data-Parallel Algorithm for Elevation Color Genera...

引用

2nd International Conference on Mechanical Engineering and Green Manufacturing (MEGM 2012)

作者： Zhou, Zhengkang Wang, Jian Zhao, Huacheng Xi An Jiao Tong Univ Dept Comp Sci & Technol Xian 710049 Peoples R China

ISBN: (纸本)9783037853771

With the ever increasing resolution of scanned elevations, the efficiency of terrain rendering is still unsatisfactory in three-dimensional visualization. In order to improve the speed of terrain rendering, we had studied the root cause of the undesirable visualization efficiency in Visual Terrain Project, and found that one of the performance limitations was elevation color generation step. Therefore, we proposed a modified algorithm on the basis of data parallel mode, which can be well performed on GPU. Experimental results show that application performance is raised significantly by data parallel algorithm, and the improved algorithm can obtain satisfactory simulation results via CUDA.

关键词： algorithm elevation color generation data parallel CUDA terrain rendering

来源：评论

学校读者我要写书评

暂无评论

parallel functional programming on recursively defined data via data-parallel recursion

引用

Journal of Functional Programming 1999年第4期9卷 427-462页

作者： Nishimura, Susumu Ohori, Atsushi Res. Inst. for Mathematical Sciences Kyoto University Sakyo-ku Kyoto 606-8502 Japan

This article proposes a new language mechanism for data-parallel processing of dynamically allocated recursively defined data. Different from the conventional array-based data-parallelism, it allows parallel processing of general recursively defined data such as lists or trees in a functional way. This is achieved by representing a recursively defined datum as a system of equations, and defining new language constructs for parallel transformation of a system of equations. By integrating them with a higher-order functional language, we obtain a functional programming language suitable for describing data-parallel algorithms on recursively defined data in a declarative way. The language has an ML style polymorphic type system and a type sound operational semantics that uniformly integrates the parallel evaluation mechanism with the semantics of a typed functional language. We also show the intended parallel execution model behind the formal semantics, assuming an idealized distributed memory multicomputer.

关键词： trees Functional programming semiotics Integrate Semantics distributed memory data parallel parallel PROCESSING (COMPUTERS) Language multicomputer

来源：评论

学校读者我要写书评

暂无评论

Energy-Efficient Execution of data-parallel Applications on Heterogeneous Mobile Platforms 33

Energy-Efficient Execution of Data-Parallel Applications on ...

引用

33rd IEEE International Conference on Computer Design (ICCD)

作者： Prakash, Alok Wang, Siqi Irimiea, Alexandru Eugen Mitra, Tulika Natl Univ Singapore Sch Comp Singapore 117417 Singapore

ISBN: (纸本)9781467371667

State-of-the-art mobile system-on-chips (SoC) include heterogeneity in various forms for accelerated and energy-efficient execution of diverse range of applications. The modern SoCs now include programmable cores such as CPU and GPU with very different functionality. The SoCs also integrate performance heterogeneous cores with different power-performance characteristics but the same instruction-set architecture such as ARM big. LITTLE. In this paper, we first explore and establish the combined benefits of functional heterogeneity and performance heterogeneity in improving power-performance behavior of data parallel applications. Next, given an application specified in OpenCL, we present a static partitioning strategy to execute the application kernel across CPU and GPU cores along with voltage-frequency setting for individual cores so as to obtain the best power-performance tradeoff. We achieve over 19% runtime improvement by exploiting the functional and performance heterogeneities concurrently. In addition, energy saving of 36% is achieved by using appropriate voltage-frequency setting without significantly degrading the runtime improvement from concurrent execution.

关键词： Graphics processing units Kernel Mobile communication Multicore processing Pipelines Runtime Runtime mobile communication pipeline Graphics Processing Unit data parallel Frequency setting Kernel central processing units Energy efficiency moving platform piping core Core samples Seeds parallel execution

来源：评论

学校读者我要写书评

暂无评论

Simultaneous Finite Automata: An Efficient data-parallel Model for Regular Expression Matching

Simultaneous Finite Automata: An Efficient Data-Parallel Mod...

引用

42nd Annual International Conference on parallel Processing (ICPP)

作者： Sin'ya, Ryoma Matsuzaki, Kiminori Sassa, Masataka Tokyo Inst Technol Dept Math & Comp Sci Tokyo Japan Kochi Univ Technol Sch Informat Kochi Japan

ISBN: (纸本)9780769551173

Automata play important roles in wide area of computing and the growth of multicores calls for their efficient parallel implementation. Though it is known in theory that we can perform the computation of a finite automaton in parallel by simulating transitions, its implementation has a large overhead due to the simulation. In this paper we propose a new automaton called simultaneous finite automaton (SFA) for efficient parallel computation of an automaton. The key idea is to extend an automaton so that it involves the simulation of transitions. Since an SFA itself has a good property of parallelism, we can develop easily a parallel implementation without overheads. We have implemented a regular expression matcher based on SFA, and it has achieved over 10-times speedups on an environment with dual hexa-core CPUs in a typical case.

关键词： Automata data parallel Regular Expression Syntactic Monoid

来源：评论

学校读者我要写书评

暂无评论

Transparent CPU-GPU Collaboration for data-parallel Kernels on Heterogeneous Systems

Transparent CPU-GPU Collaboration for Data-Parallel Kernels ...

引用

22nd International Conference on parallel Architectures and Compilation Techniques (PACT)

作者： Lee, Janghaeng Samadi, Mehrzad Park, Yongjun Mahlke, Scott Univ Michigan Adv Comp Architecture Lab Ann Arbor MI 48109 USA

ISBN: (纸本)9781479910212

Heterogeneous computing on CPUs and GPUs has traditionally used fixed roles for each device: the GPU handles data parallel work by taking advantage of its massive number of cores while the CPU handles non data-parallel work, such as the sequential code or data transfer management. Unfortunately, this work distribution can be a poor solution as it under utilizes the CPU, has difficulty generalizing beyond the single CPU-GPU combination, and may waste a large fraction of time transferring data. Further, CPUs are performance competitive with GPUs on many workloads, thus simply partitioning work based on the fixed roles may be a poor choice. In this paper, we present the single kernel multiple devices (SKMD) system, a framework that transparently orchestrates collaborative execution of a single data-parallel kernel across multiple asymmetric CPUs and GPUs. The programmer is responsible for developing a single data-parallel kernel in OpenCL, while the system automatically partitions the workload across an arbitrary set of devices, generates kernels to execute the partial workloads, and efficiently merges the partial outputs together. The goal is performance improvement by maximally utilizing all available resources to execute the kernel. SKMD handles the difficult challenges of exposed data transfer costs and the performance variations GPUs have with respect to input size. On real hardware, SKMD achieves an average speedup of 29% on a system with one multicore CPU and two asymmetric GPUs compared to a fastest device execution strategy for a set of popular OpenCL kernels.

关键词： GPGPU OpenCL Collaboration data parallel

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：