Parallel Computation Method Research Articles

A parallel tag affinity computation for social tagging systems using MapReduce

Tag affinity is the relationship between tags. It is a useful information for search and recommendation in social tagging systems. Tag affinity is measured by several types of tag cooccurrence frequency. The computation of tag affinity is a time-consuming task as the tagging information is accumulated. To alleviate this problem, we propose a parallel tag affinity computation method using MapReduce. We present MapReduce algorithms for computing three types of tag affinity measures: macro, micro, and bigram tag cooccurrence frequency. Our experimental results show that the proposed MapReduce-based approach not only significantly outperforms existing methods based on a relational database but also provides high scalability. To the best of our knowledge, this approach is the first tag affinity computation on MapReduce.

A three-dimensional domain decomposition method for large-scale DFT electronic structure calculations

With tens of petaflops supercomputers already in operation and exaflops machines expected to appear within the next 10 years, efficient parallel computational methods are required to take advantage of such extreme-scale machines. In this paper, we present a three-dimensional domain decomposition scheme for enabling large-scale electronic structure calculations based on density functional theory (DFT) on massively parallel computers. It is composed of two methods: (i) the atom decomposition method and (ii) the grid decomposition method. In the former method, we develop a modified recursive bisection method based on the moment of inertia tensor to reorder the atoms along a principal axis so that atoms that are close in real space are also close on the axis to ensure data locality. The atoms are then divided into sub-domains depending on their projections onto the principal axis in a balanced way among the processes. In the latter method, we define four data structures for the partitioning of grid points that are carefully constructed to make data locality consistent with that of the clustered atoms for minimizing data communications between the processes. We also propose a decomposition method for solving the Poisson equation using the three-dimensional FFT in Hartree potential calculation, which is shown to be better in terms of communication efficiency than a previously proposed parallelization method based on a two-dimensional decomposition. For evaluation, we perform benchmark calculations with our open-source DFT code, OpenMX, paying particular attention to the O(N) Krylov subspace method. The results show that our scheme exhibits good strong and weak scaling properties, with the parallel efficiency at 131,072 cores being 67.7% compared to the baseline of 16,384 cores with 131,072 atoms of the diamond structure on the K computer.

Research on the Application of Computer Virtual Image Technology in Artistic Photography

The principles of using computer parallel calculation method carries out in-depth analysis and research for computer art photography virtual imaging technology, and carries out image processing experiments for the artistic image rendering, we can get better art photography image processing effect. Combined with photography image virtual image mathematical model, this paper designs a computer program of photography image art rendering, and carries out art rendering for a landscape painting, we can get ideal art modification effect. Finally, this paper begins to calculated results for the parallel computing of art photography rendering, the calculation is found that the parallel computing is less time-consuming than general algorithm and with the increase of pixel time-consuming increased gently, it doesnt appear large fluctuations. For an image of the same resolution to carry on data processing, the general algorithm takes 55 seconds to complete the task. For parallel computing, it needs to3 seconds to complete, greatly saving time and computer resources, and providing the theory reference for the development of artistic photographys virtual image technology

A Parallel Computing Method of Social Tag Cooccurrence Relation

Cooccurrence relation extraction of abundant tags is very significant f or information retrieve and tag recommendation. This paper presents a parallel extraction method of tag cooccurrence relatio n based on cloud computing. At last, two experiments show that the parallel extraction method can improve the performance of retrievals greatly and have good scalability. The research may play a role in the facilitative effect on the expanding application of social tagging s ystem with abundant, multiplex, complex annotated data.

Study on Parallel Computing Method of 3-D Dynamic Analysis of Large-Scale Structure

In order to solve 3-D dynamic problem for large-scale structure, based on HDDM (Hierarchical Domain Decomposition Method) method, 3-D dynamic finite element parallel computing method of large-scale structure is studied. In this paper, BDD method and Newmark-β average acceleration method in time integration is used. Based on ADVENTRUE program, the corresponding linear implicit transient dynamic parallel program which is based on the MPI parallel environment and can be used in shared memory and distributed memory parallel computer is prepared. Several pre-conditioners are contrasted using some examples, and the results show that BDD preconditioner or BDD-DIAG preconditioner can be used in 3-D dynamic analysis for large-scale structure.

Study on Parallel Computing Method of 3-D Dynamic Analysis of Large-Scale Structure

In order to solve 3-D dynamic problem for large-scale structure, based on HDDM (Hierarchical Domain Decomposition Method) method, 3-D dynamic finite element parallel computing method of large-scale structure is studied. In this paper, BDD method and Newmark-β average acceleration method in time integration is used. Based on ADVENTRUE program, the corresponding linear implicit transient dynamic parallel program which is based on the MPI parallel environment and can be used in shared memory and distributed memory parallel computer is prepared. Several pre-conditioners are contrasted using some examples, and the results show that BDD preconditioner or BDD-DIAG preconditioner can be used in 3-D dynamic analysis for large-scale structure.

Measurement of the latency parameters of the Multi-BSP model: a multicore benchmarking approach

Computer benchmarking is a common method for measuring the parameters of a computational model. It helps to measure the parameters of any computer. With the emergence of multicore computers, the evaluation of computers was brought under consideration. Since these types of computers can be viewed and considered as parallel computers, the evaluation methods for parallel computers may be appropriate for multicore computers. However, because multicore architectures seriously focus on cache hierarchy, there is a need for new and different benchmarks to evaluate them correctly. To this end, this paper presents a method for measuring the parameters of one of the most famous multicore computational models, namely Multi-Bulk Synchronous Parallel (Multi-BSP). This method measures the hardware latency parameters of multicore computers, namely communication latency (g i ) and synchronization latency (L i ) for all levels of the cache memory hierarchy in a bottom-up manner. By determining the parameters, the performance of algorithms on multicore architectures can be evaluated as a sequence.

Application of High Performance Parallel Computing Based on GPU

This paper introduces parallel feature of the GPU, which will help GPU parallel computation methods to achieve the parallelization of PSO parallel path search process; and reduce the increasingly high problem of PSO (PSO: Particle Swarm Optimization) in time and space complexity. The experimental results show: comparing with CPU mode, GPU platform calculation improves the search rate and shortens the calculation time.

An Optimal Software Framework for Parallel Computation of CRC

CRC is a common error detection method used in different areas such as information storage and data communication. CRC depends on modulo-2 division by a predetermined divisor called the generator. In this method, the transmitter divides the message by the generator and concatenates the calculated residue to the message. CRC is not able to detect every kind of errors. The properties of the generator determine the range of errors which are detectable in the receiver side. The division operation is currently performed sequentially, so developing methods for parallel computation of the residue makes CRC suitable for network protocols and software applications. This paper presents a novel software framework for parallel computation of CRC using ODP polynomials.

A High-performance temporal-spatial discretization method for the parallel computing of river basins

The distributed basin model (DBM) has become one of the most effective tools in river basin studies. In order to overcome the efficiency bottleneck of DBM, an effective parallel-computing method, named temporal-spatial discretization method (TSDM), is proposed. In space, TSDM adopts the sub-basin partitioning manner to the river basin. Compared to the existing sub-basin-based parallel methods, more computable units can be supplied, organized and dispatched using TSDM. Through the characteristic of the temporal-spatial dual discretization, TSDM is capable of exploiting the river-basin parallelization degree to the maximum extent and obtaining higher computing performance. A mathematical formula assessing the maximum speedup ratio (MSR) of TSDM is provided as well. TSDM is independent of the implementation of any physical models and is preliminarily tested in the Lhasa River basin with 1-year rainfall-runoff process simulated. The MSR acquired in the existing traditional way is 7.98. Comparatively, the MSR using TSDM equals to 15.04 under the present limited computing resources, which appears to still have potential to keep increasing. The final results demonstrate the effectiveness and applicability of TSDM.

A Parallel Computation Method for Heuristic Attribute Reduction Using Reduced Decision Tables

In this paper, we propose a parallel computation framework for a heuristic attribute reduction method. Attribute reduction is a key technique to use rough set theory as a tool in data mining. The authors have previously proposed a heuristic attribute reduction method to compute as many relative reducts as possible from a given dataset with numerous attributes. We parallelize our method by using open multiprocessing. We also evaluate the performance of a parallelized attribute reduction method by experiments.

Review of Monte Carlo modeling of light transport in tissues

A general survey is provided on the capability of Monte Carlo (MC) modeling in tissue optics while paying special attention to the recent progress in the development of methods for speeding up MC simulations. The principles of MC modeling for the simulation of light transport in tissues, which includes the general procedure of tracking an individual photon packet, common light-tissue interactions that can be simulated, frequently used tissue models, common contact/noncontact illumination and detection setups, and the treatment of time-resolved and frequency-domain optical measurements, are briefly described to help interested readers achieve a quick start. Following that, a variety of methods for speeding up MC simulations, which includes scaling methods, perturbation methods, hybrid methods, variance reduction techniques, parallel computation, and special methods for fluorescence simulations, as well as their respective advantages and disadvantages are discussed. Then the applications of MC methods in tissue optics, laser Doppler flowmetry, photodynamic therapy, optical coherence tomography, and diffuse optical tomography are briefly surveyed. Finally, the potential directions for the future development of the MC method in tissue optics are discussed.

An Improved Approach to the PageRank Problems

We introduce a partition of the web pages particularly suited to the PageRank problems in which the web link graph has a nested block structure. Based on the partition of the web pages, dangling nodes, common nodes, and general nodes, the hyperlink matrix can be reordered to be a more simple block structure. Then based on the parallel computation method, we propose an algorithm for the PageRank problems. In this algorithm, the dimension of the linear system becomes smaller, and the vector for general nodes in each block can be calculated separately in every iteration. Numerical experiments show that this approach speeds up the computation of PageRank.

Parallel approximation algorithms for minimum routing cost spanning tree

With the popularity of internet, more and more data is transferred in the network. With the enormous information, the network bandwidth is still a bottleneck. The internet must provide high quality of service to ensure that the information can be transferred fluently. There are some common factors that can affect the quality of services, such as delay time, building cost, routing cost, loss probability, and bandwidth. Estimation of some factors, such as building cost, can be solved in polynomial time, but the estimations of other factors, such as finding the minimum routing cost spanning tree MRCT, are NP-hard problems. In this paper, we focus on improving two MRCT approximation algorithms 2 and 15/8 approximation with parallel-computing methods and obtain the impressive experiment results with reduced calculation time.

Three-dimensional numerical simulation of the single-channel pulsed-power vacuum device

A single-channel pulsed-power vacuum device is studied. Based on the platform of CHIPIC, the limitation of memory and computation speeds of a single computer is broken through by the method of parallel computation of multi-computers. For the model of the whole device, the models of the parts of feed-in, convergence and transmission in the single-channel pulsed-power device are built, and the parameters are set up. The obtained voltages between the cathode and anode, as well as the currents of the cathode and the anode using the proposed models are in accordance with the theoretical ones. The single-channel pulsed-power vacuum device also can retain magnetically insulated and achieve the function of convergence of power. This shows the feasibility of the single-channel pulsed-power vacuum device.

A Novel Method of Parallel Computation for the Whole Scene Test in Power System

In this paper, design and implementation of a new parallel computing method based on CUDA (Compute Unified Device Architecture) platform is described in detail. The method includes algorithm of matrix fraction and partial LU decomposition that are used to support parallel computing for simulation of the whole scene test in power system. The paper describes all the steps of algorithm implementation for deploying the software with CUDA technology. And performances are evaluated by test, in which two Jacobi matrix equations of power flow calculation with different dimension are selected as the test case. The results are compared with that of serial computing and MPI (Message Passing Interface). The results show that the method proposed in this paper has better performance of acceleration on CUDA platform, which can support the simulation of the whole scene test on wide-area power grid.

An Efficient Parallel Computing Method for the Processing of Large Sensed Data

In recent years we witness the advent of the Internet of Things and the wide deployment of sensors in many applications for collecting and aggregating data. Efficient techniques are required to analyze these massive data for supporting intelligent decisions making. Partial differential problems which involve large data are the most common in the engineering and scientific research. For simulations of large-scale three-dimensional partial differential equations, the intensive computation ability and large amounts of memory requirements for modeling are the main research problems. To address the two challenges, this paper provided an effective parallel method for partial differential equations. The proposed approach combines the overlapping domain decomposition strategy and the multi-core cluster technology to achieve parallel simulations of partial differential equations, uses the finite difference method to discretize equations and adopts the hybrid MPI/OpenMP programming model to exploit two-level parallelism on a multi-core cluster. The three-dimensional groundwater flow model with the parallel finite difference overlapping domain decomposition strategy was successfully set up and carried out by the parallel MPI/OpenMP implementation on a multi-core cluster with two nodes. The experimental results show that the proposed parallel approach can efficiently simulate partial differential problems with large amounts of data.

A Hybrid Optimization Based on GPU Parallel Computing Method and its Application of Three Dimensional Large Eddy Simulations

Using compute Unified device architecture (CUDA), a traditional computational fluid dynamics (CFD) program is paralleled and optimized based on graphic processing unit (GPU). The calculation process is divided into two parts as serial and parallel. Their main characteristics are analyzed and different optimization schemes are given. CPU (central processing unit) and GPU work respectively as flow control and high-speed parallel computation. Bandwidth between devices is applied effectively. Data transfer between devices is moderately improved to simplify algorithm. Finally, the method is verified by simulating a three-dimensional isotropic homogeneous turbulence flow field. The calculation uses large eddy simulation (LES) method with secondary filter and solves the three-dimensional N-S equations. The maximum grid number achieves 8,000,000 and takes 33 seconds each step. All calculations are using ordinary single desktop computer, optimized acceleration ratio can reach 9.

Study on Hardware Parallel Computing Method of Wavelet Transform Based on FIR filter

Study on Hardware Parallel Computing Method of Wavelet Transform Based on FIR filter

Parallel calculation methods for molecular weight distribution of batch free radical polymerization

The molecular weight distribution (MWD) is at the core of establishing key quality indices for free radical polymerization processes. Due to large-scale features of the rigorous model, consisting of a very large number of differential and algebraic equations, dynamic simulation of MWD is always challenging. A sequential variable decoupling method (SVD) has been proposed to calculate the MWD for any reasonably large chain-length number. In the current paper, parallel computing methods were developed to accelerate the MWD simulation. Both coarse-grained and fine-grained parallelism methods have been proposed. A theoretical analysis of the proposed methods was conducted to demonstrate its high efficiency in parallel computing. Both Intel multi-core-processor-based and NVIDIA graphics-processing-unit-based parallel computing platforms were implemented, achieving significant speedups in computation for MWD simulation.