A top-level view of the architecture of the Parallel Programming Support Environment (PPSE), which is an experimental set of tools for use in the design and implementation of software systems for parallel computing sy...
详细信息
A top-level view of the architecture of the Parallel Programming Support Environment (PPSE), which is an experimental set of tools for use in the design and implementation of software systems for parallel computing systems, is proposed. The tools include a graphical design editor, a graphical target machine description system, a task mapper/scheduler tool, parallel code generator, and graphical aids for performance analysis. The objective is, to the extent possible, to design and develop parallel software with little regard for the details of the architecture of the target machine, programming language, or parallel computing paradigm that the program is to use. A prototype of the environment running on the Macintosh was developed. It was used to parallelize a small image-processing application and several simple programs.< >
The ability of multilayer perceptrons (MLPs) trained with backpropagation to classify vowels excised from natural continuous speech is examined. Two spectral representations are compared: spectrograms and cochleagrams...
详细信息
The ability of multilayer perceptrons (MLPs) trained with backpropagation to classify vowels excised from natural continuous speech is examined. Two spectral representations are compared: spectrograms and cochleagrams. The features used to train the MLPs include discrete Fourier transform (DFT) or cochleagram coefficients from a single frame in the middle of the vowel, or coefficients from each third of the vowel. The effects of estimates of pitch, duration, and the relative amplitude of the vowel were investigated. The experiments show that with coefficients alone, the cochleagram is superior to the spectrogram in classification performance for all experimental conditions. With the three additional features, however, the results are comparable. Perceptual experiments with trained human listeners on the same data revealed that MLPs perform much better than humans on vowels excised from context.< >
This paper discusses studies on the performance of a parallel iterative algorithm implemented on an array of transputers connected in a mesh configuration. The iterative algorithm under consideration is the finite dif...
详细信息
This paper discusses studies on the performance of a parallel iterative algorithm implemented on an array of transputers connected in a mesh configuration. The iterative algorithm under consideration is the finite difference method for the solution of partial differential equations. Analytical expressions for the execution times of the various steps of the algorithm are derived by studying its computation and communication characteristics. These expressions are validated by comparing the theoretical results of the performance with the experimental values obtained on a transputer array. Then the analytical model is used to estimate the performance of the algorithm for varying number of transputers in the array and for varying grid sizes. An important objective of this paper is to study the influence of the convergence detection overhead on the performance of the algorithm. We present an approach to minimize the overhead. Convergence detection is one of the dominant factors that affects the performance of the algorithm, since it involves a substantial amount of computation and communication. In order to reduce this overhead, the proposed algorithm checks convergence once in every certain number of iterations, k c . The method of determining an optimal value of k c is given. Further, the time taken for convergence detection is estimated for the best case, worst case, and average case situations.
We propose a new strategy to recognize the maximum subcube in an n-cube multiprocessor. This subcube recognition algorithm can be done in both serial and parallel and analyzed. This strategy will enhance the performan...
ISBN:
(纸本)0897913418
We propose a new strategy to recognize the maximum subcube in an n-cube multiprocessor. This subcube recognition algorithm can be done in both serial and parallel and analyzed. This strategy will enhance the performance drastically so that our algorithm will outperform the buddy system by a factor nCk, the gray strategy by nCk/2 and Al-Dhelaan [8] by nCk/(k(n-k)+1) in cube recognition. We present a very efficient processor allocation strategy which makes larger contiguous spaces for the new coming job than buddy, gray strategy and Al-Dhelaan [8] do. Furthermore, this new strategy is suitable for static as well as dynamic processors allocation and it results in a less fragmentation and higher fault tolerance. Also we describe an efficient procedure for task migration under the new strategy: 1) goal configuration under the new strategy 2) node-mapping between source and destination node 3) the shortest deadlock-free routing algorithm.
A description is given of two known strategies for static-processor allocation in an n-cube multiprocessor, namely, the buddy-system strategy and the gray-code strategy. A strategy is proposed that outperforms the fir...
详细信息
A description is given of two known strategies for static-processor allocation in an n-cube multiprocessor, namely, the buddy-system strategy and the gray-code strategy. A strategy is proposed that outperforms the first by (n-k+1) and the second by (n-k+1)/2 in a cube recognition. The authors' strategy is suitable for static as well as dynamic processor allocation, and it results in less system fragmentation, more subcube recognition, and higher fault tolerance. The authors also introduce an extension to their strategy that will enhance the performance drastically so that their algorithm together with the extension will outperform the buddy system by a factor of (k(n-k)+1) and the gray strategy by (k(n-k)+1)/2 in cube recognition. The implementation details for these algorithms are described.< >
The cube-connected cycles network is discussed as a cube whose 2k vertices are cycles of k nodes each. It is suitable for realization for VLSI since it satisfies the properties of degree boundedness of nodes (=3). Bro...
详细信息
The cube-connected cycles network is discussed as a cube whose 2k vertices are cycles of k nodes each. It is suitable for realization for VLSI since it satisfies the properties of degree boundedness of nodes (=3). Broadcasting is a procedure by which a processor can pass a message to all other processors in the network nonredundantly. This is extremely important for diagnosis of the network, distribution agreement or clock synchronization. A simple yet efficient algorithm is developed for broadcasting in the cube-connected cycles network. Another broadcasting algorithm for the cube-connected cycles network in the presence of some faulty processors is also developed.< >
作者：
JOLLIFF, JVCALLAHAN, CMUSNCapt. James V. Jolliff
USNgraduated from the U. S. Naval Academy in 1954. Following graduation he served in the USS S. N. Moore (DD—747) and USS Cimarron (AO—22). He received his MS degrees in Naval Architecture from Webb Institute of Naval Architecture and in Financial Management from The George Washington University. He culminated his education at The Catholic University of America where he was awarded his Doctorate in Ocean Engineering in 1972. Capt. Jolliff has served in Naval Shipyards as Ship Superintendent Assistant Repair Officer and Assistant Planning & Estimating Superintendent and as such was primarily concerned with the repair and conversion of U. S. Navy skips. In addition he has served as Maintenance Officer Staff of Commander Mine Force U. S. Pacific Fleet as Co—Chairman of the Naval Engineering Division
Engineering Department U. S. Naval Academy and as CV Design Manager in the Advanced Concepts Division and as Head
Ship Survivability Office Naval Ship Engineering Center. An active member of ASNE since 1966 he has served as a member of the National Council and is currently the Chairman of the Journal Committee. He has had several papers presented at ASNE Day and published in the Journal and in 1976 was one of the recipients of the ASNE President's Award. At the present time he is assigned as the Commanding Officer Naval Coastal Systems Laboratory (NCSL) Panama City Fla. Mr. Casville M. Callahanis a native of Southwest Virginia where he attended Elementary and Secondary School prior to his three year's service in the U. S. Navy during World War II. He graduated from Lincoln Memorial University
Harrogate Tenn. in 1950 receiving his BS degree in Mathematics. In 1952 he received his MS degree in Mathematics from Auburn University Auburn Ala. and taught mathematics at Lincoln Memorial University and at Florida State University Tallahassee Fla. prior to joining the staff of the Mine Defense Laboratory in 1955. He has progressed through a variety of assignments as the Labo
Test and Evaluation have become paramount in today's department of Defense acquisition process. Therefore, the U. S. Navy requires both private and public facilities to accomplish the final goals of the “Fly befo...
Test and Evaluation have become paramount in today's department of Defense acquisition process. Therefore, the U. S. Navy requires both private and public facilities to accomplish the final goals of the “Fly before Buy” concept. Such a facility exists at the Naval Coastal Systems Laboratory (NCSL); an integral part of the Chief of Naval Material's, Director of Navy Laboratories organization. This paper briefly addresses the Laboratory, its mission, and its history. This is followed by an in—depth facilities overview in order to create an understanding of the slow but steady evolution of NCSL's unique fixed facilities. These facilities, when coupled to the local natural environment, provide a unique in situ test and evaluation capability which is unequalled in the United States for assessing seagoing coastal systems. Of prime consideration is the Range Date Acquisition Center (RADAC) and Its ancillary subsystems for tracking, environmental monitoring, communications, and post run analysis. The paper is concluded with a discussion of both past and present use of the aforementioned facilities with an emphasis on user acceptance and future potential growth.
This book gathers the proceedings of the Seventh International Conference on Computational science and Technology 2020 (ICCST 2020), held in Pattaya, Thailand, on 29–30 August 2020. The respective contributions offer...
详细信息
ISBN:
(数字)9789813340695
ISBN:
(纸本)9789813340688;9789813340718
This book gathers the proceedings of the Seventh International Conference on Computational science and Technology 2020 (ICCST 2020), held in Pattaya, Thailand, on 29–30 August 2020. The respective contributions offer practitioners and researchers a range of new computational techniques and solutions, identify emerging issues, and outline future research directions, while also showing them how to apply the latest large-scale, high-performance computational methods.
暂无评论