检索结果-内蒙古大学图书馆

30th International Conference on parallel Architectures and Compilation Techniques (PACT)

作者： Kim, Joonsung Hur, Suyeon Lee, Eunbok Lee, Seungho Kim, Jangwoo Seoul Natl Univ Dept Elect & Comp Engn Seoul South Korea

ISBN: (纸本)9781665442787

Emerging natural language processing (NLP) models have become more complex and bigger to provide more sophisticated NLP services. Accordingly, there is also a strong demand for scalable and flexible computer infrastructure to support these large-scale, complex, and diverse NLP models. However, existing proposals cannot provide enough scalability and flexibility as they neither identify nor optimize a wide spectrum of performance-critical operations appearing in recent NLP models and only focus on optimizing specific operations. In this paper, we propose NLP-Fast, a novel system solution to accelerate a wide spectrum of large-scale NLP models. NLP-Fast mainly consists of two parts: (1) NLP-Perf : an in-depth performance analysis tool to identify critical operations in emerging NLP models and (2) NLP-Opt: three end-to-end optimization techniques to accelerate the identified performance-critical operations on various hardware platforms (e.g., CPU, GPU, FPGA). In this way, NLP-Fast can accelerate various types of NLP models on different hardware platforms by identifying their critical operations through NLP-Perf and applying the NLP-Opt's holistic optimizations. We evaluate NLP-Fast on CPU, GPU, and FPGA, and the overall throughputs are increased by up to 2.92x, 1.59x, and 4.47x over each platform's baseline. We release NLP-Fast to the community so that users are easily able to conduct the NLP-Fast's analysis and apply NLP-Fast's optimizations for their own NLP applications.

关键词： Architecture Computation/Dataflow Optimization Natural Language Processing (NLP) parallel algorithm

来源：评论

学校读者我要写书评

暂无评论

Improved memory management for solving neutron transport via a novel Modular Ray Tracing (MRT) approach embedded in parallel method of characteristic (MOC) framework

引用

PROGRESS IN NUCLEAR ENERGY 2021年 132卷 103590-103590页

作者： Porhemmat, Mohammadhadi Mahzoon, Mojtaba Salehi, Ali Akbar Hadad, Kamal Hosseinipour, Khalil Shiraz Univ Sch Mech Engn Shiraz Iran Sharif Univ Dept Energy Engn Tehran Iran

Poor memory management can be considered as the first restrictive barrier to the implementation of the neutron transport equation for large reactors. The current research presents a new MRT with the ability to reduce the required memory and increase cache coherency. Also, given the high impact of the angular quadrature set used in the transport calculation, the effects of the angular set are investigated. As a reliable trial for an angular quadrature set examination, comparisons between the Forward and Adjoint transport calculations are performed. Three types of angular quadrature sets are adopted for our optimized MOC steady-state calculation. Based on two known constraints for the quadrature set evaluation, the accuracy and computational cost of the presented sets are examined. Satisfying these two conditions, our implemented Legendre-Chebyshev quadrature set with S20 angular order results in 7.0000E-06 and the 9.0000E-06 difference with reference Keff of G5G7 benchmark in the forward and adjoint calculation respectively. As well, our optimized parallel algorithm for MOC results in 6.18 speedup with 16 threads.

关键词： Modular ray tracing Forward and adjoint neutron transport MOC Kernel parallel algorithm

来源：评论

学校读者我要写书评

暂无评论

Research of the Mechanism of External Hormonal Regulation of the Development of Phyto- and Zooplankton Populations Using Supercomputer Technologies 15th

Research of the Mechanism of External Hormonal Regulation of...

引用

15th International Scientific Conference on parallel Computational Technologies (PCT)

作者： Sukhinov, Alexander, I Chistyakov, Alexander E. Belova, Yulia Epifanov, Alexander Nikitina, Alla, V Don State Tech Univ Rostov Na Donu Russia Supercomp & Neurocomp Res Ctr Taganrog Russia Southern Fed Univ Rostov Na Donu Russia

ISBN: (纸本)9783030816919;9783030816902

The paper is devoted to modeling the dynamics of phyto- and zooplankton populations in coastal systems, taking into account the mechanism of external hormonal regulation and relying on NVIDIA Tesla K80 graphics accelerators. The mathematical model of the dynamics of plankton populations is based on a system of non-stationary equations of convection-diffusion-reaction with nonlinear terms. We carry out the linearization of the continuous problem and construct its discrete analog by splitting the original three-dimensional problem into a two-dimensional problem and a one-dimensional problem, in conformity with the linearized model. To construct the discrete two-dimensional model, we use a linear combination of the Upwind and the Standard Leapfrog difference schemes, which makes it possible to increase the modeling accuracy in the case of large values of the grid Peclet number. We introduce into the model of ectocrine regulation of interacting phytoplankton species a nonlinear dependence between the phytoplankton growth rate and the concentration of metabolites. By doing so, we manage to describe the ability of algal metabolites to control their growth, even under conditions of an excess of nutrients, which corresponds to modern concepts on the functioning of hydrobiocenosis. The solution to the problem is numerically implemented on a graphics accelerator for massively parallel computations, which significantly reduces the operating time of the software module. We study a parallel computing model on a GPU on a sequence of thickening grids with different numbers of threads.

关键词： Plankton populations External hormonal regulation Coastal system parallel algorithm Software package GPU

来源：评论

学校读者我要写书评

暂无评论

A 10000-Image-per-Second parallel algorithm for Real-Time Detection of MARFEs on JET

引用

IEEE TRANSACTIONS ON PLASMA SCIENCE 2013年第2期41卷 341-349页

作者： de Albuquerque, Marcio Portes Murari, Andrea Giovani, M. Alves, Nilton, Jr. de Albuquerque, Marcelo Portes Ctr Brasileiro Pesquisas Fis BR-22299180 Rio De Janeiro Brazil Italian Natl Agcy New Technol Energy & Sustainabl European Atom Energy Community Consorzio Reversed Field EXpt Assoc I-35127 Padua Italy

This paper presents a very high-speed image processing algorithm applied to multi-faceted asymmetric radiation from the edge (MARFE) detection on the Joint European Torus. The algorithm was built in serial and parallel versions and written in C/C+ using OpenCV, cvBlob, and LibSVM libraries. The code implemented was characterized by its accuracy and run-time performance. The final result of the parallel version achieves a correct detection rate of 97.6% for MARFE identification and an image processing rate of more than 10 000 frame per second. The parallel version divides the image processing chain into two groups and seven tasks. One group is responsible for Background Image Estimation and Image Binarization modules, and the other is responsible for region Feature Extraction and Pattern Classification. At the same time and to maximize the workload distribution, the parallel code uses data parallelism and pipeline strategies for these two groups, respectively. A master thread is responsible for opening, signaling, and transferring images between both groups. The algorithm has been tested in a dedicated Intel symmetric-multiprocessing computer architecture with a Linux operating system.

关键词： Image processing multi-faceted asymmetric radiation from the edge detection multi-faceted asymmetric radiation from the edge identification parallel algorithm

来源：评论

学校读者我要写书评

暂无评论

Compact Local IRBF and Domain Decomposition Method for solving PDEs using a Distributed termination detection based parallel algorithm

引用

CMES-COMPUTER MODELING IN ENGINEERING & SCIENCES 2013年第1期92卷 1-31页

作者： Pham-Sy, N. Tran, C. -D. Hoang-Trieu, T. -T. Mai-Duy, N. Tran-Cong, T. Univ So Queensland Fac Engn & Surveying Computat Engn & Sci Res Ctr Toowoomba Qld 4350 Australia

Compact Local Integrated Radial Basis Function (CLIRBF) methods based on Cartesian grids can be effective numerical methods for solving partial differential equations (PDEs) for fluid flow problems. The combination of the domain decomposition method and function approximation using CLIRBF methods yields an effective coarse-grained parallel processing approach. This approach has enabled not only each sub-domain in the original analysis domain to be discretised by a separate CLIRBF network but also compact local stencils to be independently treated. The present algorithm, namely parallel CLIRBF, achieves higher throughput in solving large scale problems by, firstly, parallel processing of sub-regions which constitute the original domain and, secondly, accelerating the convergence rate within each sub-region using groups of CLIRBF stencils in which function approximations are carried out by parallel processes. The procedure is illustrated with several numerical examples of PDEs and lid-driven cavity problem using Message Passing Interface supported by MATLAB.

关键词： Integrated RBFs Compact Local Stencils Domain Decomposition Method parallel algorithm Distributed Termination Detection

来源：评论

学校读者我要写书评

暂无评论

parallel random matrix particle swarm optimization scheduling algorithms with budget constraints on cloud computing systems

引用

APPLIED SOFT COMPUTING 2021年第PartA期113卷 107914-107914页

作者： Tang, Xiaoyong Shi, Cheng Deng, Tan Wu, Zhiqiang Yang, Li Changsha Univ Sci & Technol Sch Comp & Commun Engn Changsha 410114 Peoples R China

Nowadays, increasing number of Internet of Things and mobile Internet application services are migrated to cloud computing systems. One of the most important cloud challenges for this business is to optimize services cost. The efficient way to deal with this challenge is to improve the performance of resource management and task scheduling in cloud systems. However, this kind of service scheduling is a typical combinatorial optimization problem, and it is difficult to obtain the optimal solution. In this study, we first formalize this cloud services on virtual machines (VMs) with budget constraints scheduling problem. Then, we propose a random matrix particle swarm optimization scheduling algorithm (RMPSO), which uses the random integer matrix to represent its position and a feasible task scheduling scheme, to achieve the optimal total cost of cloud services. However, the drawback of this solution for large-scale systems is its high time complexity. Therefore, we propose two parallel RMPSO algorithms: CPU parallel algorithm (M-RMPSO) on multi-core system with shared memory and manycore GPU-accelerated strategy (G-RMPSO) to reduce its time complexity. Finally, the rigorous performance evaluation results clearly show that our proposed G-RMPSO outperforms M-RMPSO and existing FMPSO, HYBRID (MPSO+MCSO) in terms of cloud services total cost and algorithm execution time. Therefore, our proposed GPU-accelerated G-RMPSO algorithm is very suitable for cloud service scheduling. (C) 2021 Elsevier B.V. All rights reserved.

关键词： Cloud computing systems Particle swarm optimization parallel algorithm Task scheduling Cost

来源：评论

学校读者我要写书评

暂无评论

Efficient parallel GCD algorithms for Multicore Shared Memory Architectures 20

Efficient Parallel GCD Algorithms for Multicore Shared Memor...

引用

20th International Conference on Advances in ICT for Emerging Regions (ICTer)

作者： Pathirana, Gihan Tharaka Sotheeswaran, Sittampalam Ratnarajah, Nagulan Eastern Univ Fac Sci Dept Math Chenkaladi Sri Lanka Univ Jaffna Fac Appl Sci Dept Phys Sci Vavuniya Campus Jaffna Sri Lanka

ISBN: (纸本)9781728186535

The greatest common divisor (GCD) is used for numerous applications in number theory, modular arithmetic, encryption algorithms such as RSA, random number generation, and solving linear Diophantine equations. High-performance algorithms, which are efficiently and accurately find the GCD of two integers and n (>2) integers, are needed in the modern world for many applications in science and engineering. parallel hardware and parallel programming solve such emerging challenges that are not possible in a serial world. Modern desktop and laptop computers are equipped with multicore processors with shared memory architecture. In this paper, we develop novel efficient parallel.. integers GCD algorithms for multicore shared memory architecture. The brute force, divide-and-conquer, linear recursive and finding minimum first techniques are adopted in our novel algorithms to reduce the size and the complexity of the n integers GCD problem. Various working models of OpenMP, such as the thread-centric, loop-centric and task-centric models are utilized, which promised a more natural way of exploiting and expressing regular and irregular algorithms. A comprehensive performance analysis applies to prove the efficiency of the proposed algorithms.

关键词： GCD parallel algorithm Shared Memory Multicore OpenMP

来源：评论

学校读者我要写书评

暂无评论

Distributed Model Predictive Control for Optimal Consensus of Constrained Multi-agent Systems

Distributed Model Predictive Control for Optimal Consensus o...

引用

第42届中国控制会议

作者： Nan Bai Qishao Wang Zhisheng Duan Department of Mechanics and Engineering Science Peking University Department of Dynamics and Control Beihang University

This paper develops a framework based on distributed model predictive control to solve the optimal consensus problem for constrained multi-agent systems. Taking both transient performance and final consensus state into consideration, the optimization problem in each prediction horizon is formed as a coupled optimization problem containing a non-separable cost function with constraints. An efficient distributed algorithm is proposed with the distributed convergence conditions on the auxiliary parameters, which makes each agent solve its subproblems in parallel iterations, enhancing the running efficiency of the algorithm. The stability of the closed-loop system is analyzed, providing distributed stability conditions that only depend on the local information of each agent. Numerical simulations verify the validity of theoretical results by utilizing the proposed approach to a multi-robot system.

关键词： Distributed Model Predictive Control Distributed Optimization Multi-agent System parallel algorithm

来源：评论

学校读者我要写书评

暂无评论

Positive Semidefinite Programming: Mixed, parallel, and Width-Independent 2020

Positive Semidefinite Programming: Mixed, Parallel, and Widt...

引用

52nd Annual ACM SIGACT Symposium on Theory of Computing (STOC)

作者： Jambulapati, Arun Lee, Yin Tat Li, Jerry Padmanabhan, Swati Tian, Kevin Stanford Univ Stanford CA 94305 USA Univ Washington Seattle WA 98195 USA Microsoft Res Seattle WA USA Microsoft Res Redmond WA USA

ISBN: (纸本)9781450369794

We study mixed packing and covering semidefinite programs (SDPs) and give the first approximation algorithm with polylogarithmic dependence on width. Mixed packing and covering SDPs constitute a fundamental algorithmic primitive with applications in combinatorial optimization, robust learning, and quantum complexity. The current approximate solvers for positive semidefinite programming can handle only pure packing instances, and technical hurdles prevent their generalization to a wider class of positive instances. For a given multiplicative accuracy of c, our algorithm takes O (log(3)(ndp) . epsilon(-3)) parallelizable iterations, where n, d are the problem dimensions and rho is a width parameter of the instance, generalizing or improving all previous parallel algorithms in the positive linear and semidefinite programming literature. When specialized to pure packing SDPs, our algorithm's iteration complexity is O (log(2) (nd) . epsilon(-2)), a slight improvement and derandomization of the state-of-the-art [AI,016, rrl16, WNIMI-215]. For several common structured instances, the iterations of our algorithm run in nearly-linear time. In doing so, we give matrix analytic techniques to overcome obstacles that have stymied prior approaches to this problem, as stated in [PTZ16, NIRWZ16]. Crucial to our analysis are a simplification of existing algorithms for mixed positive linear programs, achieved by removing an asymmetry from modifying covering constraints, and a suite of matrix inequalities with proofs based on analyzing the Schur complements of matrices in a higher dimension. We hope that our algorithm and techniques open the door to improved solvers for positive semidefinite programming and its applications.

关键词： semidefinite programming mixed packing and covering approximation algorithm parallel algorithm width-independent algorithm

来源：评论

学校读者我要写书评

暂无评论

Research on parallel Method of Moments On Domestic Supercomputer Platform

Research on Parallel Method of Moments On Domestic Supercomp...

引用

IEEE-MTT-S International Conference on Numerical Electromagnetic and Multiphysics Modeling and Optimization (NEMO)

作者： Chen, Yan Gu, Zongjing Li, Yanyan Shao, Xuehui Sun, Peng Qiang, Xiaomin Yang, Baisheng Beijing Aerosp Automat Control Inst Natl Key Lab Sciense & Technol Aerosp Intelligenc Beijing Peoples R China China Elect Technol Instruments Co Ltd Qingdao Peoples R China

ISBN: (纸本)9781728169668

The parallel method of moment is firstly studied based on many-core supercomputer platform with a peak computing speed of more than 100PFlops, which is the first supercomputer of China built by all use independent technology. For the matrix solution process with intensive computation and high CPU utilization, the matrix multiplication operation with high concurrency degree is executed at the slave with strong computing power by using the xMath mathematical library of "Sunway TaiHu Light" system. The parallel efficiency of solving matrix is up to 90% using one-way communication LU algorithm named CALU, shows better scalability than open ScaLAPACK mathematical library. The numerical results show that CALU algorithm has a strong advantage in solving method of moment equations based on domestically-made supercomputer platform, which greatly speeds up the process of solving matrix.

关键词： parallel algorithm method of moment computational electromagnetics LU decomposition

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：