Search results for: parallel computation

Authors: Duygu Kan, Mehmet Cayoren

Abstract:

Sommerfeld-integrals (SIs) are commonly encountered in electromagnetics problems involving analysis of antennas and scatterers embedded in planar multilayered media. Generally speaking, the analytical solution of SIs is unavailable, and it is well known that numerical evaluation of SIs is very time consuming and computationally expensive due to the highly oscillating and slowly decaying nature of the integrands. Therefore, fast computation of SIs has a paramount importance. In this paper, a parallel code has been developed to speed up the computation of SI in the framework of calculation of dyadic Green’s function in multilayered media. OpenMP shared memory approach is used to parallelize the SI algorithm and resulted in significant time savings. Moreover accelerating the computation of dyadic Green’s function is discussed based on the parallel SI algorithm developed.

Keywords: Sommerfeld-integrals, multilayer dyadic Green’s function, OpenMP, shared memory parallel programming

Procedia PDF Downloads 214

Authors: Claude Tadonki

Abstract:

We address the issues related to the computation of the covariance matrix. This matrix is likely to be ill conditioned following its canonical expression, thus consequently raises serious numerical issues. The underlying linear system, which therefore should be solved by means of iterative approaches, becomes computationally challenging. A huge number of iterations is expected in order to reach an acceptable level of convergence, necessary to meet the required accuracy of the computation. In addition, this linear system needs to be solved at each iteration following the general form of the covariance matrix. Putting all together, its comes that we need to compute as fast as possible the associated matrix-vector product. This is our purpose in the work, where we consider and discuss skillful formulations of the problem, then propose a parallel implementation of the matrix-vector product involved. Numerical and performance oriented discussions are provided based on experimental evaluations.

Keywords: covariance-matrix, multicore, numerical computing, parallel computing

Procedia PDF Downloads 285

Authors: Dimitris Varsamis, Christos Talagkozis, Alkiviadis Tsimpiris, Paris Mastorocostas

Abstract:

The aim of this work is the parallel implementation of k-means in MATLAB, in order to reduce the execution time. Specifically, a new function in MATLAB for serial k-means algorithm is developed, which meets all the requirements for the conversion to a function in MATLAB with parallel computations. Additionally, two different variants for the definition of initial values are presented. In the sequel, the parallel approach is presented. Finally, the performance tests for the computation times respect to the numbers of features and classes are illustrated.

Keywords: K-means algorithm, clustering, parallel computations, Matlab

Procedia PDF Downloads 345

Authors: Sugeng Rianto, P.W. Arinto Yudi, Soemarno Muhammad Nurhuda

Abstract:

A computation of a 3D compressible fluid flow for virtual environment with haptic interaction can be a non-trivial issue. This is especially how to reach good performances and balancing between visualization, tactile feedback interaction, and computations. In this paper, we describe our approach of computation methods based on parallel programming on a GPU. The 3D fluid flow solvers have been developed for smoke dispersion simulation by using combinations of the cubic interpolated propagation (CIP) based fluid flow solvers and the advantages of the parallelism and programmability of the GPU. The fluid flow solver is generated in the GPU-CPU message passing scheme to get rapid development of haptic feedback modes for fluid dynamic data. A rapid solution in fluid flow solvers is developed by applying cubic interpolated propagation (CIP) fluid flow solvers. From this scheme, multiphase fluid flow equations can be solved simultaneously. To get more acceleration in the computation, the Navier-Stoke Equations (NSEs) is packed into channels of texel, where computation models are performed on pixels that can be considered to be a grid of cells. Therefore, despite of the complexity of the obstacle geometry, processing on multiple vertices and pixels can be done simultaneously in parallel. The data are also shared in global memory for CPU to control the haptic in providing kinaesthetic interaction and felling. The results show that GPU based parallel computation approaches provide effective simulation of compressible fluid flow model for real-time interaction in 3D computer graphic for PC platform. This report has shown the feasibility of a new approach of solving the compressible fluid flow equations on the GPU. The experimental tests proved that the compressible fluid flowing on various obstacles with haptic interactions on the few model obstacles can be effectively and efficiently simulated on the reasonable frame rate with a realistic visualization. These results confirm that good performances and balancing between visualization, tactile feedback interaction, and computations can be applied successfully.

Keywords: CIP, compressible fluid, GPU programming, parallel computation, real-time visualisation

Procedia PDF Downloads 404

Authors: Mustafa Elfituri, Jonathan Cook

Abstract:

Recently, graph-based computations have become more important in large-scale scientific computing as they can provide a methodology to model many types of relations between independent objects. They are being actively used in fields as varied as biology, social networks, cybersecurity, and computer networks. At the same time, graph problems have some properties such as irregularity and poor locality that make their performance different than regular applications performance. Therefore, parallelizing graph algorithms is a hard and challenging task. Initial evidence is that standard computer architectures do not perform very well on graph algorithms. Little is known exactly what causes this. The Graph500 benchmark is a representative application for parallel graph-based computations, which have highly irregular data access and are driven more by traversing connected data than by computation. In this paper, we present results from analyzing the performance of various example implementations of Graph500, including a shared memory (OpenMP) version, a distributed (MPI) version, and a hybrid version. We measured and analyzed all the factors that affect its performance in order to identify possible changes that would improve its performance. Results are discussed in relation to what factors contribute to performance degradation.

Keywords: graph computation, graph500 benchmark, parallel architectures, parallel programming, workload characterization.

Procedia PDF Downloads 112

Authors: Kuniyoshi Abe

Abstract:

Bi-conjugate gradient (Bi-CG) is a well-known method for solving linear equations Ax = b, for x, where A is a given n-by-n matrix, and b is a given n-vector. Typically, the dimension of the linear equation is high and the matrix is sparse. A number of hybrid Bi-CG methods such as conjugate gradient squared (CGS), Bi-CG stabilized (Bi-CGSTAB), BiCGStab2, and BiCGstab(l) have been developed to improve the convergence of Bi-CG. Bi-CGSTAB has been most often used for efficiently solving the linear equation, but we have seen the convergence behavior with a long stagnation phase. In such cases, it is important to have Bi-CG coefficients that are as accurate as possible, and the stabilization strategy, which stabilizes the computation of the Bi-CG coefficients, has been proposed. It may avoid stagnation and lead to faster computation. Motivated by a large number of processors in present petascale high-performance computing hardware, the scalability of Krylov subspace methods on parallel computers has recently become increasingly prominent. The main bottleneck for efficient parallelization is the inner products which require a global reduction. The resulting global synchronization phases cause communication overhead on parallel computers. The parallel variants of Krylov subspace methods reducing the number of global communication phases and hiding the communication latency have been proposed. However, the numerical stability, specifically, the convergence speed of the parallel variants of Bi-CGSTAB may become worse than that of the standard Bi-CGSTAB. In this paper, therefore, we compare the convergence speed between the standard Bi-CGSTAB and the parallel variants by numerical experiments and show that the convergence speed of the standard Bi-CGSTAB is faster than the parallel variants. Moreover, we propose the stabilization strategy for the parallel variants.

Keywords: bi-conjugate gradient stabilized method, convergence speed, Krylov subspace methods, linear equations, parallel variant

Procedia PDF Downloads 138

Authors: Jian Zheng, Xinhua Ni, Xiequan Liu

Abstract:

According to isotropy of parallel fiber eutectic, the no- damage strain field in parallel fiber eutectic is obtained from the flexibility tensor of parallel fiber eutectic. Considering the damage behavior of parallel fiber eutectic, damage variables are introduced to determine the strain field of parallel fiber eutectic. The damage strains in the matrix, interphase, and fiber of parallel fiber eutectic are quantitatively analyzed. Results show that damage strains are not only associated with the fiber volume fraction of parallel fiber eutectic, but also with the damage degree.

Keywords: damage strain, initial strain, fiber volume fraction, parallel fiber eutectic

Procedia PDF Downloads 537

Authors: Sheng-Po Tseng, Che-Hua Yang

Abstract:

Motivation and Objective: Ultrasonic guided waves become an important tool for nondestructive evaluation of structures and components. Guided waves are used for the purpose of identifying defects or evaluating material properties in a nondestructive way. While guided waves are applied for evaluating material properties, instead of knowing the properties directly, preliminary signals such as time domain signals or frequency domain spectra are first revealed. With the measured ultrasound data, inversion calculation can be further employed to obtain the desired mechanical properties. Methods: This research is development of high speed inversion calculation technique for obtaining full-field mechanical properties from the quantitative laser ultrasound visualization system (QLUVS). The quantitative laser ultrasound visualization system (QLUVS) employs a mirror-controlled scanning pulsed laser to generate guided acoustic waves traveling in a two-dimensional target. Guided waves are detected with a piezoelectric transducer located at a fixed location. With a gyro-scanning of the generation source, the QLUVS has the advantage of fast, full-field, and quantitative inspection. Results and Discussions: This research introduces two important tools to improve the computation efficiency. Firstly, graphic procession unit (GPU) with large amount of cores are introduced. Furthermore, combining the CPU and GPU cores, parallel procession scheme is developed for the inversion of full-field mechanical properties based on the QLUVS data. The newly developed inversion scheme is applied to investigate the computation efficiency for single-layered and double-layered plate-like samples. The computation efficiency is shown to be 80 times faster than unparalleled computation scheme. Conclusions: This research demonstrates a high-speed inversion technique for the characterization of full-field material properties based on quantitative laser ultrasound visualization system. Significant computation efficiency is shown, however not reaching the limit yet. Further improvement can be reached by improving the parallel computation. Utilizing the development of the full-field mechanical property inspection technology, full-field mechanical property measured by non-destructive, high-speed and high-precision measurements can be obtained in qualitative and quantitative results. The developed high speed computation scheme is ready for applications where full-field mechanical properties are needed in a nondestructive and nearly real-time way.

Keywords: guided waves, material characterization, nondestructive evaluation, parallel processing

Procedia PDF Downloads 175

Authors: Trung Hieu Tran, Jesse O'Hanley, Russell Fowler

Abstract:

When in-wheel wireless charging technology for electric vehicles becomes mature, a need for such integrated charging stations network development is essential. In this paper, we thus investigate the optimisation problem of in-wheel wireless electric vehicle charging network design. A mixed-integer linear programming model is formulated to solve into optimality the problem. In addition, a meta-heuristic algorithm is proposed for efficiently solving large-sized instances within a reasonable computation time. A parallel computing strategy is integrated into the algorithm to speed up its computation time. Experimental results carried out on the benchmark instances show that our model and algorithm can find the optimal solutions and their potential for practical applications.

Keywords: electric vehicle, wireless charging station, mathematical programming, meta-heuristic algorithm, parallel computing

Procedia PDF Downloads 50

Authors: Zhao Bin, Yan Dan Lei

Abstract:

In this paper, a multi-pattern parallel PRBS generator and a dedicated parallel BER tester is proposed for the 8-Gbps On-chip interconnection testing. A unique full-parallel PRBS checker is also proposed. The proposed design, together with the custom-designed high-speed parallel-to-serial and the serial-to-parallel circuit, will be used to test different on-chip interconnection transceivers. The design is implemented in TSMC 28nm CMOS technology with working voltage at 1.0 V. The serial to parallel ratio is 8:1 so the parallel PRBS generation and BER Tester can be run at lower speed.

Keywords: PRBS, BER, high speed, generator

Procedia PDF Downloads 684

Authors: Jie Yang, Jian-Min Meng

Abstract:

A GPU-based floating object detection scheme is presented in this paper which is designed for floating mine detection tasks. This system uses contrast and motion information to eliminate as many false positives as possible while avoiding false negatives. The GPU computation platform is deployed to allow detecting objects in real-time. From the experimental results, it is shown that with certain configuration, the GPU-based scheme can speed up the computation up to one thousand times compared to the CPU-based scheme.

Keywords: object detection, GPU, motion estimation, parallel processing

Procedia PDF Downloads 446

Authors: Seongsoo Lee

Abstract:

Motion estimation occupies the heaviest computation in HEVC (high efficiency video coding). Many fast algorithms such as TZS (test zone search) have been proposed to reduce the computation. Still the huge computation of the motion estimation is a critical issue in the implementation of HEVC video codec. In this paper, motion estimator architecture with optimized number of PEs (processing element) is presented by exploiting early termination. It also reduces hardware size by exploiting parallel processing. The presented motion estimator architecture has 8 PEs, and it can efficiently perform TZS with very high utilization of PEs.

Keywords: motion estimation, test zone search, high efficiency video coding, processing element, optimization

Procedia PDF Downloads 333

Authors: Ming-Yang Guo, Cheng-Xian Wu, Wei-Xiang Chen, Chun-Yuan Lin, Yen-Jen Lin, Ann-Shyn Chiang

Abstract:

With more and more Drosophila Driver and Neuron images, it is an important work to find the similarity relationships among them as the functional inference. There is a general problem that how to find a Drosophila Driver image, which can cover a set of Drosophila Driver/Neuron images. In order to solve this problem, the intersection/union region for a set of images should be computed at first, then a comparison work is used to calculate the similarities between the region and other images. In this paper, three encoding schemes, namely Integer, Boolean, Decimal, are proposed to encode each image as a one-dimensional structure. Then, the intersection/union region from these images can be computed by using the compare operations, Boolean operators and lookup table method. Finally, the comparison work is done as the union region computation, and the similarity score can be calculated by the definition of Tanimoto coefficient. The above methods for the region computation are also implemented in the multi-core CPUs environment with the OpenMP. From the experimental results, in the encoding phase, the performance by the Boolean scheme is the best than that by others; in the region computation phase, the performance by Decimal is the best when the number of images is large. The speedup ratio can achieve 12 based on 16 CPUs. This work was supported by the Ministry of Science and Technology under the grant MOST 106-2221-E-182-070.

Keywords: Drosophila driver image, Drosophila neuron images, intersection/union computation, parallel processing, OpenMP

Procedia PDF Downloads 204

Authors: H. Samadiyeh, R. Khajavi

Abstract:

A new FD-based procedure, modal finite difference method (MFDM), is proposed for seismic wave propagation modeling, in which simulation is dealt with in the modal space. The method employs eigenvalues of a characteristic matrix formed by appropriate time-space FD stencils. Since MFD runs for different modes are totally independent of each other, MFDM can easily be parallelized while considerable simplicity in parallel-algorithm is also achieved. There is no requirement to any domain-decomposition procedure and inter-core data exchange. More important is the possibility to skip processing of less-significant modes, which enables one to adjust the procedure up to the level of accuracy needed. Thus, in addition to considerable ease of parallel programming, computation and storage costs are significantly reduced. The method is qualified for its efficiency by some numerical examples.

Keywords: Finite Difference Method, Graphics Processing Unit (GPU), Message Passing Interface (MPI), Modal, Wave propagation

Procedia PDF Downloads 269

Authors: Chunye Gong, Ming Tie, Jie Liu, Weimin Bao, Xinbiao Gan, Shengguo Li, Bo Yang, Xuguang Chen, Tiaojie Xiao, Yang Sun

Abstract:

Large-scale CFD simulation relies on high-performance parallel computing, and the load balance is the key role which affects the parallel efficiency. This paper focuses on the load-balancing problem of parallel CFD simulation with structured mesh. A mathematical model for this load-balancing problem is presented. The genetic algorithm, fitness computing, two-level code are designed. Optimal selector, robust operator, and local optimization operator are designed. The properties of the presented genetic algorithm are discussed in-depth. The effects of optimal selector, robust operator, and local optimization operator are proved by experiments. The experimental results of different test sets, DLR-F4, and aircraft design applications show the presented load-balancing algorithm is robust, quickly converged, and is useful in real engineering problems.

Keywords: genetic algorithm, load-balancing algorithm, optimal variation, local optimization

Procedia PDF Downloads 140

Authors: Nebras Gharbi, Itebeddine Ghorbel

Abstract:

In the latest years and with the advancements of the multicore computing world, the constraint programming community tried to benefit from the capacity of new machines and make the best use of them through several parallel schemes for constraint solving. In this paper, we propose a survey of the different proposed approaches to solve Constraint Satisfaction Problems using parallel architectures. These approaches use in a different way a parallel architecture: the problem itself could be solved differently by several solvers or could be split over solvers.

Keywords: constraint programming, parallel programming, constraint satisfaction problem, speed-up

Procedia PDF Downloads 288

Authors: Abdullah Bal, Sevdenur Bal

Abstract:

This paper purposes a new type of hardware application for training of cellular neural networks (CNN) using optical joint transform correlation (JTC) architecture for image feature extraction. CNNs require much more computation during the training stage compare to test process. Since optoelectronic hardware applications offer possibility of parallel high speed processing capability for 2D data processing applications, CNN training algorithm can be realized using Fourier optics technique. JTC employs lens and CCD cameras with laser beam that realize 2D matrix multiplication and summation in the light speed. Therefore, in the each iteration of training, JTC carries more computation burden inherently and the rest of mathematical computation realized digitally. The bipolar data is encoded by phase and summation of correlation operations is realized using multi-object input joint images. Overlapping properties of JTC are then utilized for summation of two cross-correlations which provide less computation possibility for training stage. Phase-only JTC does not require data rearrangement, electronic pre-calculation and strict system alignment. The proposed system can be incorporated simultaneously with various optical image processing or optical pattern recognition techniques just in the same optical system.

Keywords: CNN training, image processing, joint transform correlation, optoelectronic hardware

Procedia PDF Downloads 475

Authors: Adarsh Shroff

Abstract:

Big data is a collection of dataset so large and complex that it becomes difficult to process using data base management tools. To perform operations like search, analysis, visualization on big data by using data mining; which is the process of extraction of patterns or knowledge from large data set. In recent years, the data mining applications become stale and obsolete over time. Incremental processing is a promising approach to refreshing mining results. It utilizes previously saved states to avoid the expense of re-computation from scratch. This project uses i2MapReduce, an incremental processing extension to Map Reduce, the most widely used framework for mining big data. I2MapReduce performs key-value pair level incremental processing rather than task level re-computation, supports not only one-step computation but also more sophisticated iterative computation, which is widely used in data mining applications, and incorporates a set of novel techniques to reduce I/O overhead for accessing preserved fine-grain computation states. To optimize the mining results, evaluate i2MapReduce using a one-step algorithm and three iterative algorithms with diverse computation characteristics for efficient mining.

Keywords: big data, map reduce, incremental processing, iterative computation

Procedia PDF Downloads 315

Authors: Muhammad Waqar Aziz, Syed Abdul Baqi Shah

Abstract:

The execution time analysis is fundamental to the successful design and execution of real-time embedded software. In such analysis, the Worst-Case Execution Time (WCET) of a program is a key measure, on the basis of which system tasks are scheduled. The WCET analysis of embedded software is also needed for system understanding and to guarantee its behavior. WCET analysis can be performed statically (without executing the program) or dynamically (through measurement). Traditionally, research on the WCET analysis assumes sequential code running on single-core platforms. However, as computation is steadily moving towards using a combination of parallel programs and multi-core hardware, new challenges in WCET analysis need to be addressed. In this article, we report our experiences of performing the WCET analysis of Parallel Embedded Software (PES) running on multi-core platform. The primary purpose was to investigate how WCET estimates of PES can be computed statically, and how they can be derived dynamically. Our experiences, as reported in this article, include the challenges we faced, possible suggestions to these challenges and the workarounds that were developed. This article also provides observations on the benefits and drawbacks of deriving the WCET estimates using the said methods and provides useful recommendations for further research in this area.

Keywords: embedded software, worst-case execution-time analysis, static flow analysis, measurement-based analysis, parallel computing

Procedia PDF Downloads 299

Authors: Yogita M. Ahire, Nedal M. Mohammed, Ahmed A. Hamoud

Abstract:

Scientific computation outsourcing is gaining popularity because it allows customers with limited computing resources and storage devices to outsource complex computation workloads to more powerful service providers. However, it raises some security and privacy concerns and challenges, such as customer input and output privacy, as well as cloud cheating behaviors. This study was motivated by these concerns and focused on privacy-preserving Two-Point Boundary Value Problems (BVP) as a common and realistic instance for verifiable safe multiparty computing. We'll look at the safe and verifiable schema with correctness guarantees by utilizing standard multiparty approaches to compute the result of a computation and then solely using verifiable ways to check that the result was right.

Keywords: verifiable computing, cloud computing, secure and privacy BVP, secure computation outsourcing

Procedia PDF Downloads 63

Authors: Zhen Peng, Baifeng Wu

Abstract:

Modern computing technology enters the era of parallel computing with the trend of sustainable and scalable parallelism. Single Instruction Multiple Data (SIMD) is an important way to go along with the trend. It is able to gather more and more computing ability by increasing the number of processor cores without the need of modifying the program. Meanwhile, in the field of scientific computing and engineering design, many computation intensive applications are facing the challenge of increasingly large amount of data. Data parallel computing will be an important way to further improve the performance of these applications. In this paper, we take the accurate collision detection in building information modeling as an example. We demonstrate a model for constructing a data parallel algorithm. According to the model, a complex object is decomposed into the sets of simple objects; collision detection among complex objects is converted into those among simple objects. The resulting algorithm is a typical SIMD algorithm, and its advantages in parallelism and scalability is unparalleled in respect to the traditional algorithms.

Keywords: data parallelism, collision detection, single instruction multiple data, building information modeling, continuous scalability

Procedia PDF Downloads 258

Authors: Nicolò Vaiana, Giorgio Serino

Abstract:

In this paper, a one-dimensional (1d) Parallel Elasto- Plastic Model (PEPM), able to simulate the uniaxial dynamic behavior of seismic isolators having a continuously decreasing tangent stiffness with increasing displacement, is presented. The parallel modeling concept is applied to discretize the continuously decreasing tangent stiffness function, thus allowing to simulate the dynamic behavior of seismic isolation bearings by putting linear elastic and nonlinear elastic-perfectly plastic elements in parallel. The mathematical model has been validated by comparing the experimental force-displacement hysteresis loops, obtained testing a helical wire rope isolator and a recycled rubber-fiber reinforced bearing, with those predicted numerically. Good agreement between the simulated and experimental results shows that the proposed model can be an effective numerical tool to predict the forcedisplacement relationship of seismic isolators within relatively large displacements. Compared to the widely used Bouc-Wen model, the proposed one allows to avoid the numerical solution of a first order ordinary nonlinear differential equation for each time step of a nonlinear time history analysis, thus reducing the computation effort, and requires the evaluation of only three model parameters from experimental tests, namely the initial tangent stiffness, the asymptotic tangent stiffness, and a parameter defining the transition from the initial to the asymptotic tangent stiffness.

Keywords: base isolation, earthquake engineering, parallel elasto-plastic model, seismic isolators, softening hysteresis loops

Procedia PDF Downloads 255

Authors: Rizwan Rizwan

Abstract:

Secure Multiparty Computation is an emerging area of research within the cryptographic community, enabling secure collaboration among multiple parties while safeguarding their sensitive data. Secure Multiparty Computation has been extensively studied in the context of plane geometry, its application to complex solid geometry shapes remains relatively unexplored. This research paper aims to bridge this gap by proposing a solution for the secure multiparty computation of intersecting tetrahedra. We present an approach to calculate the volume of intersecting tetrahedra securely while preserving the privacy of the input data provided by each participating party. The proposed solution leverages accepted simulation paradigms to prove the privacy of the computation. We thoroughly analyze the computational and communication complexities of our approach, demonstrating that they closely align with the minimum theoretical complexity for the problems at hand. This optimal nature of our solution ensures efficient and secure collaborative geometric computations.

Keywords: cryptography, secure multiparty computation, solid geometry, protocol, simulation paradigm

Procedia PDF Downloads 25

Authors: Sun Lim, Kyun Jung

Abstract:

In this paper, we describe the control strategy of high speed parallel robot system with EtherCAT network. This work deals the parallel robot system with centralized control on the real-time operating system such as window TwinCAT3. Most control scheme and algorithm is implemented master platform on the PC, the input and output interface is ported on the slave side. The data is transferred by maximum 20usecond with 1000byte. EtherCAT is very high speed and stable industrial network. The control strategy with EtherCAT is very useful and robust on Ethernet network environment. The developed parallel robot is controlled pre-design nonlinear controller for 6G/0.43 cycle time of pick and place motion tracking. The experiment shows the good design and validation of the controller.

Keywords: parallel robot control, etherCAT, nonlinear control, parallel robot inverse kinematic

Procedia PDF Downloads 541

Authors: Orhun Vural, Oguz Bayat, Rustu Akay, Osman N. Ucan

Abstract:

This study presents a new parallel approach clustering of GPS data. Evaluation has been made by comparing execution time of various clustering algorithms on GPS data. This paper aims to propose a parallel based on neighborhood K-means algorithm to make it faster. The proposed parallelization approach assumes that each GPS data represents a vehicle and to communicate between vehicles close to each other after vehicles are clustered. This parallelization approach has been examined on different sized continuously changing GPS data and compared with serial K-means algorithm and other serial clustering algorithms. The results demonstrated that proposed parallel K-means algorithm has been shown to work much faster than other clustering algorithms.

Keywords: parallel k-means algorithm, parallel clustering, clustering algorithms, clustering on flowing data

Procedia PDF Downloads 188

Authors: Sergey Kopysov, Nikita Nedozhogin, Leonid Tonkov

Abstract:

The article presents a parallel iterative solver for large sparse linear systems which can be used on a heterogeneous platform. Traditionally, the problem of solving linear systems does not scale well on multi-CPU/multi-GPUs clusters. For example, most of the attempts to implement the classical conjugate gradient method were at best counted in the same amount of time as the problem was enlarged. The paper proposes the pipelined variant of the conjugate gradient method (PCG), a formulation that is potentially better suited for hybrid CPU/GPU computing since it requires only one synchronization point per one iteration instead of two for standard CG. The standard and pipelined CG methods need the vector entries generated by the current GPU and other GPUs for matrix-vector products. So the communication between GPUs becomes a major performance bottleneck on multi GPU cluster. The article presents an approach to minimize the communications between parallel parts of algorithms. Additionally, computation and communication can be overlapped to reduce the impact of data exchange. Using the pipelined version of the CG method with one synchronization point, the possibility of asynchronous calculations and communications, load balancing between the CPU and GPU for solving the large linear systems allows for scalability. The algorithm is implemented with the combined use of technologies: MPI, OpenMP, and CUDA. We show that almost optimum speed up on 8-CPU/2GPU may be reached (relatively to a one GPU execution). The parallelized solver achieves a speedup of up to 5.49 times on 16 NVIDIA Tesla GPUs, as compared to one GPU.

Keywords: conjugate gradient, GPU, parallel programming, pipelined algorithm

Procedia PDF Downloads 130

Authors: Wen-Bao Qiao, Jean-Charles Créput

Abstract:

To accelerate the solution for large scale traveling salesman problems (TSP), a parallel 2-opt local search algorithm with simple implementation based on Graphics Processing Unit (GPU) is presented and tested in this paper. The parallel scheme is based on technique of data decomposition by dynamically assigning multiple K processors on the integral tour to treat K edges’ 2-opt local optimization simultaneously on independent sub-tours, where K can be user-defined or have a function relationship with input size N. We implement this algorithm with doubly linked list on GPU. The implementation only requires O(N) memory. We compare this parallel 2-opt local optimization against sequential exhaustive 2-opt search along integral tour on TSP instances from TSPLIB with more than 10000 cities.

Keywords: parallel 2-opt, double links, large scale TSP, GPU

Procedia PDF Downloads 594

Authors: Malihe Aliasgari, Yousef Nejatbakhsh

Abstract:

The era of Big Data and the immensity of real-life datasets compels computation tasks to be performed in a distributed fashion, where the data is dispersed among many servers that operate in parallel. However, massive parallelization leads to computational bottlenecks due to faulty servers and stragglers. Stragglers refer to a few slow or delay-prone processors that can bottleneck the entire computation because one has to wait for all the parallel nodes to finish. The problem of straggling processors, has been well studied in the context of distributed computing. Recently, it has been pointed out that, for the important case of linear functions, it is possible to improve over repetition strategies in terms of the tradeoff between performance and latency by carrying out linear precoding of the data prior to processing. The key idea is that, by employing suitable linear codes operating over fractions of the original data, a function may be completed as soon as enough number of processors, depending on the minimum distance of the code, have completed their operations. The problem of matrix-matrix multiplication in the presence of practically big sized of data sets faced with computational and memory related difficulties, which makes such operations are carried out using distributed computing platforms. In this work, we study the problem of distributed matrix-matrix multiplication W = XY under storage constraints, i.e., when each server is allowed to store a fixed fraction of each of the matrices X and Y, which is a fundamental building of many science and engineering fields such as machine learning, image and signal processing, wireless communication, optimization. Non-secure and secure matrix multiplication are studied. We want to study the setup, in which the identity of the matrix of interest should be kept private from the workers and then obtain the recovery threshold of the colluding model, that is, the number of workers that need to complete their task before the master server can recover the product W. The problem of secure and private distributed matrix multiplication W = XY which the matrix X is confidential, while matrix Y is selected in a private manner from a library of public matrices. We present the best currently known trade-off between communication load and recovery threshold. On the other words, we design an achievable PSGPD scheme for any arbitrary privacy level by trivially concatenating a robust PIR scheme for arbitrary colluding workers and private databases and the proposed SGPD code that provides a smaller computational complexity at the workers.

Keywords: coded distributed computation, private information retrieval, secret sharing, stragglers

Procedia PDF Downloads 91

Authors: Xie Kefeng, Zhang He

Abstract:

For the stability and control demand of offshore small floating platform, a 2-HUS/U parallel mechanism was presented as offshore platform. Inverse kinematics was obtained by institutional constraint equation, and the dynamic model of offshore 2-HUS/U parallel platform was derived based on rigid body’s Lagrangian method. The equivalent moment of inertia, damping and driving force/torque variation of offshore 2-HUS/U parallel platform were analyzed. A numerical example shows that, for parallel platform of given motion, system’s equivalent inertia changes 1.25 times maximally. During the movement of platform, they change dramatically with the system configuration and have coupling characteristics. The maximum equivalent drive torque is 800 N. At the same time, the curve of platform’s driving force/torque is smooth and has good sine features. The control system needs to be adjusted according to kinetic equation during stability and control and it provides a basis for the optimization of control system.

Keywords: 2-HUS/U platform, dynamics, Lagrange, parallel platform

Procedia PDF Downloads 319

Authors: G. Khamooshian

Abstract:

One of the main points of application of the mechanisms of the series and parallel is the subject of managing them. The control of this mechanism and similar mechanisms is one that has always been the intention of the scholars. On the other hand, modeling the behavior of the system is difficult due to the large number of its parameters, and it leads to complex equations that are difficult to solve and eventually difficult to control. In this paper, a six-linkage robot has been presented that could be used in different areas such as medical robots. Using these robots needs a robust control. In this paper, the system equations are first found, and then the system conversion function is written. A new controller has been designed for this robot which could be used in other parallel robots and could be very useful. Parallel robots are so important in robotics because of their stability, so methods for control of them are important and the robust controller, especially in parallel robots, makes a sense.

Keywords: 3-RRS, 6 linkage, parallel robot, control

Procedia PDF Downloads 122