Efficient Parallel Algorithms Research Articles

A parallel implementation of fluid–solid interaction solver using an immersed boundary method

A parallel scheme is developed to solve fluid–solid interaction problems using the immersed boundary method (IBM). The fluid solver is based on a non-staggered pressure-based finite volume method, targeted to solving 3D incompressible low Reynolds number flows. Domain decomposition is employed to partition the global mesh into local meshes that are distributed over processors. A new efficient parallel marking algorithm is introduced to track the fluid–solid interface for IBM implementation. A second layer of ghost cells is included in the interpolation in order to produce consistent solutions between serial and parallel runs. Three linear solvers are implemented: (1) Algebraic Multi Grid (AMG), (2) BCGSTAB with AMG preconditioner, and (3) BCGSTAB with ILU0 preconditioner. The parallel performance and benchmarks of the linear solvers are investigated on different architectures by solving the 2D diffusion–convection problem and the 3D flow over sphere problem. Finally, the parallel implementation is validated by a micro-cantilever damping case.

A parallel MPI + OpenMP + OpenCL algorithm for hybrid supercomputations of incompressible flows

The work is devoted to the development of efficient parallel algorithms for large-scale simulations of incompressible flows on hybrid supercomputers based on massively-parallel accelerators. The governing equations are discretized using a high-order finite-volume scheme for Cartesian staggered meshes with the only restriction that, at least, one direction is periodic. Its “classical” MPI+OpenMP parallel implementation for CPUs was designed to scale till 100,000 CPU cores. The new hybrid algorithm is developed on a base of a multi-level parallel model that exploits several layers of parallelism of a modern hybrid supercomputer. In this model, MPI and OpenMP are used on the first two levels to couple nodes of a supercomputer and to engage its CPU cores. Then, computing accelerators are further used by means of the hardware independent OpenCL computing standard. In this way, the implementation is adapted to a general computing model with central processors and math co-processors. In this paper the work is focused on adapting the basic operations of the algorithm to architectures of Graphics Processing Units (GPU) without considering the multi-GPU communication scheme. Technology of porting the code to OpenCL is described, certain optimization approaches are presented and relevant performance results obtaining up to 80–90 GFLOPS on a GPU accelerator are demonstrated.Moreover, the experience with different GPU architectures is summarized and a comparison based on the particular application is given for AMD and NVIDIA GPUs as well as for CUDA and OpenCL frameworks.

Visual exploration of data by using multidimensional scaling on multicore CPU, GPU, and MPI cluster

ABSTRACTVisual and interactive data exploration requires fast and reliable tools for embedding of an original data space in 3(2)‐dimensional Euclidean space. Multidimensional scaling (MDS) is a good candidate. However, owing to at least O(M2) memory and time complexity, MDS is computationally demanding for interactive visualization of data sets consisting of order of 104 objects on computer systems, ranging from PC with multicore CPU processor, graphics processing unit (GPU) board to midrange MPI clusters. To explore interactively data sets of that size, we have developed novel efficient parallel algorithms for MDS mapping based on virtual particle dynamics. We demonstrate that the performance of our MDS algorithms implemented in compute unified device architecture environment on a PC equipped with a modern GPU board (Tesla M2090, GeForce GTX 480) is considerably faster than its MPI/OpenMP parallel implementation on the modern midrange professional cluster (10 nodes, each equipped with 2x Intel Xeon X5670 CPUs). We also show that the hybridized two‐level MPI/CUDA implementation, run on a cluster of GPU nodes, can additionally provide a linear speedup. Copyright 2013 John Wiley & Sons, Ltd.

Weighted sum-rate maximization for multi-user SIMO multiple access channels in cognitive radio networks

In this article, an efficient distributed and parallel algorithm is proposed to maximize the sum-rate and optimize the input distribution policy for the multi-user single input multiple output multiple access channel (MU-SIMO MAC) system with concurrent access within a cognitive radio (CR) network. The single input means that every user has a single antenna and multiple output means that base station(s) has multiple antennas. The main features are: (i) the power distribution for the users is updated by using variable scale factors which effectively and efficiently maximize the objective function at each iteration; (ii) distributed and parallel computation is employed to expedite convergence of the proposed distributed algorithm; and (iii) a novel water-filling with mixed constraints is investigated, and used as a fundamental block of the proposed algorithm. Due to sufficiently exploiting the structure of the proposed model, the proposed algorithm owns fast convergence. Numerical results verify that the proposed algorithm is effective and fast convergent. Using the proposed approach, for the simulated range, the required number of iterations for convergence is two and this number is not sensitive to the increase of the number of users. This feature is quite desirable for large scale systems with dense active users. In addition, it is also worth noting that the proposed algorithm is a monotonic feasible operator to the iteration. Thus, the stop criterion for computation could be easily set up.

Efficient Serial and Parallel Algorithms for Selection of Unique Oligos in EST Databases

Obtaining unique oligos from an EST database is a problem of great importance in bioinformatics, particularly in the discovery of new genes and the mapping of the human genome. Many algorithms have been developed to find unique oligos, many of which are much less time consuming than the traditional brute force approach. An algorithm was presented by Zheng et al. (2004) which finds the solution of the unique oligos search problem efficiently. We implement this algorithm as well as several new algorithms based on some theorems included in this paper. We demonstrate how, with these new algorithms, we can obtain unique oligos much faster than with previous ones. We parallelize these new algorithms to further improve the time of finding unique oligos. All algorithms are run on ESTs obtained from a Barley EST database.

Algorithms for solving inverse geophysical problems on parallel computing systems

For solving inverse gravimetry problems, efficient stable parallel algorithms based on iterative gradient methods are proposed. For solving systems of linear algebraic equations with block-tridiagonal matrices arising in geoelectrics problems, a parallel matrix sweep algorithm, a square root method, and a conjugate gradient method with preconditioner are proposed. The algorithms are implemented numerically on a parallel computing system of the Institute of Mathematics and Mechanics (PCS-IMM), NVIDIA graphics processors, and an Intel multi-core CPU with some new computing technologies. The parallel algorithms are incorporated into a system of remote computations entitled “Specialized Web-Portal for Solving Geophysical Problems on Multiprocessor Computers.” Some problems with “quasi-model” and real data are solved.

Towards an efficient parallel raycasting of unstructured volumetric data on distributed environments

Direct volume rendering of large and unstructured datasets demands high computational power and memory bandwidth. Developing an efficient parallel algorithm requires a deep understanding of the bottlenecks involved in the solutions for this problem. In this work, we make a thorough analysis of the overhead components involved in parallel volume raycasting of unstructured grids for high-resolution images on distributed environments. This evaluation has revealed potential opportunities for performance improvements. The result is a novel approach to distributed memory raycasting that includes different acceleration techniques to enhance ray distribution, face projection, memory locality, and message exchanging, while maintaining load balance. We report the gains achieved in each phase and in the complete parallel algorithm when compared with a conventional approach.

Efficient polygonization of tree trunks modeled by convolution surfaces

We present an efficient polygonization approach for tree trunks modeled by line skeleton-based convolution surfaces. A quad-dominated non-convex bounding polyhedron is firstly created along the skeleton, which is then tetrahedralized and subdivided into the pre-defined resolution. After that, the iso-surface within each tetrahedron is extracted using marching tetrahedra. Our algorithm can generate polygons with adaptive edge lengths according to the thickness of the trunk. In addition, we present an efficient CUDA-based parallel algorithm utilizing the high parallelism of the tetrahedron subdivision, the potential field calculation, and the iso-surface extraction.

Efficient parallel coordinate descent algorithm for convex optimization problems with separable constraints: Application to distributed MPC

In this paper we propose a parallel coordinate descent algorithm for solving smooth convex optimization problems with separable constraints that may arise, e.g. in distributed model predictive control (MPC) for linear network systems. Our algorithm is based on block coordinate descent updates in parallel and has a very simple iteration. We prove (sub)linear rate of convergence for the new algorithm under standard assumptions for smooth convex optimization. Further, our algorithm uses local information and thus is suitable for distributed implementations. Moreover, it has low iteration complexity, which makes it appropriate for embedded control. An MPC scheme based on this new parallel algorithm is derived, for which every subsystem in the network can compute feasible and stabilizing control inputs using distributed and cheap computations. For ensuring stability of the MPC scheme, we use a terminal cost formulation derived from a distributed synthesis. Preliminary numerical tests show better performance for our optimization algorithm than other existing methods.

A three-dimensional sharp interface Cartesian grid method for solving high speed multi-material impact, penetration and fragmentation problems

This work presents a three-dimensional, Eulerian, sharp interface, Cartesian grid technique for simulating the response of elasto-plastic solid materials to hypervelocity impact, shocks and detonations. The mass, momentum and energy equations are solved along with evolution equations for deviatoric stress and plastic strain using a third-order finite difference scheme. Material deformation occurs with accompanying nonlinear stress wave propagation; in the Eulerian framework the boundaries of the deforming material are tracked in a sharp fashion using level-sets and the conditions on the immersed boundaries are applied by suitable modifications of a ghost fluid approach. The dilatational response of the material is modeled using the Mie-Gruneisen equation of state and the Johnson–Cook model is employed to characterize the material response due to rate-dependent plastic deformation. Details are provided on the treatment of the deviatoric stress ghost state so that physically correct boundary conditions can be applied at the material interfaces. An efficient parallel algorithm is used to handle computationally intensive three-dimensional problems. The results demonstrate the ability of the method to simulate high-speed impact, penetration and fragmentation phenomena in three dimensions.

An efficient parallel algorithm for the longest path problem in meshes

The longest path problem is the problem of finding a simple path with the maximum number of vertices in a given graph, and so far it has been solved polynomially only for a few classes of graphs. This problem generalizes the well-known Hamiltonian path problem, hence it is NP-hard in general graphs. In this paper, first we give a sequential linear-time algorithm for the longest path problem in meshes. Then based on this algorithm, we present a constant-time parallel algorithm for the problem, which can be run on every parallel machine.

The single-grid multilevel method and its applications

In this paper, we propose the single-grid multilevel (SGML) method for large-scale linear systems discretized from partial differential equations. The SGML method combines the methodologies of both the geometric and the algebraic multigrid methods. It uses the underlying geometric information from the finest grid. A simple and isotropic coarsening strategy is applied to explicitly control the complexity of the hierarchical structure, and smoothers are chosen based on the property of the model problem and the underlying grid information to complement the coarsening and maintain overall efficiency. Additionally, the underlying grid is used to design an efficient parallel algorithm in order to parallelize the SGML method. We apply the SGML method on the Poisson problem and the convection diffusion problem as examples, and we present the numerical results to demonstrate the performance of the SGML method.

Parallel conversion of satellite image information for a wind energy generation forecasting model


 
 
 This paper presents an efficient parallel algorithm for the problem of converting satellite imagery in binary files. The algorithm was designed to update at global scale the land cover information used by the WRF climate model. We present the characteristics of the implemented algorithm, as well as the results of performance analysis and comparisons between two approaches to implement the algorithm. The performance analysis shows that the implemented parallel algorithm improves substantially against the sequential algorithm that solves the problem, obtaining a linear speedup.
 
 

Efficient reconstruction of biological networks via transitive reduction on general purpose graphics processors

BackgroundTechniques for reconstruction of biological networks which are based on perturbation experiments often predict direct interactions between nodes that do not exist. Transitive reduction removes such relations if they can be explained by an indirect path of influences. The existing algorithms for transitive reduction are sequential and might suffer from too long run times for large networks. They also exhibit the anomaly that some existing direct interactions are also removed.ResultsWe develop efficient scalable parallel algorithms for transitive reduction on general purpose graphics processing units for both standard (unweighted) and weighted graphs. Edge weights are regarded as uncertainties of interactions. A direct interaction is removed only if there exists an indirect interaction path between the same nodes which is strictly more certain than the direct one. This is a refinement of the removal condition for the unweighted graphs and avoids to a great extent the erroneous elimination of direct edges.ConclusionsParallel implementations of these algorithms can achieve speed-ups of two orders of magnitude compared to their sequential counterparts. Our experiments show that: i) taking into account the edge weights improves the reconstruction quality compared to the unweighted case; ii) it is advantageous not to distinguish between positive and negative interactions since this lowers the complexity of the algorithms from NP-complete to polynomial without loss of quality.

Parallel frequent itemset mining using systolic arrays

Since extraction of frequent itemsets from a transaction database is crucial to several data mining tasks such as association rule generation, so frequent itemset mining is one of the most important concepts in data mining. One of the major problems in frequent itemset mining is the explosion of the number of results which is directly effecting on the execution time of itemset mining algorithms. To address this problem, closed itemsets have been proposed, which provides concise lossless representations of the original collection of frequent itemsets. Henceforth, the frequencies of all itemsets in the original collection can be reconstructed from the reduced collection. However, the reduction provided by this exact method is not sufficient to solve the pattern explosion problem, mainly because of high dimensional datasets which have large number of items in each transaction. Colossal itemset mining is another solution to reduce the output size which will not be useful if the set of all frequent itemsets have been required. Higher level of performance improvement can be obtained from efficient scalable parallel mining methods. In this paper we represent an efficient scalable parallel algorithm using systolic arrays to conduct mining of frequent itemsets in very large, such as high dimensional, datasets. In our algorithm, we use a bit matrix to compress the dataset and mapping the mining algorithm on the systolic arrays architecture. For this purpose, each transaction of dataset represents as a row in the bit matrix. We use this bit matrix structure to model the pattern mining as a systolic array problem. Our experimental results and performance study show that this algorithm outperforms substantially the best previously developed parallel algorithms.

An Efficient Parallel Algorithm for Mining Frequent Pattern

Extraction of frequent patterns in transaction-oriented database is crucial to several data mining tasks such as association rule generation, time series analysis, classification, etc. An Efficient Parallel algorithm for Mining frequent pattern (EPM) was proposed and Fast Distributed association rules Mining (FDM) algorithm was improved. Hash table technology was used to improve the generation efficiency of the 2nd candidate items . It also reduces the number of transactions in transaction database using Tid table technology. A master-slave model of parallel algorithm for mining association rules is designed in the algorithm to reduce the communication cost. The experimental results show that this algorithm has a high efficiency to deal with large database.

P System Implementation of Dynamic Programming Stereo

Designing parallel versions of sequential algorithms has attracted renewed attention, due to recent hardware advances, including various general-purpose multi-core and many-core processors, as well as special-purpose FPGA implementations. P systems consist of networks of autonomous cells, such that each cell transforms its input signals in accord with its symbol-rewriting rules and feeds the output results into its immediate neighbours. Inherent massive intra- and inter-cell parallelisms make P systems a prospective theoretical testbed for designing efficient parallel and parallel-sequential algorithms. This paper discusses the capabilities of P systems to implement the symmetric dynamic programming stereo (SDPS) matching algorithm, which explicitly accounts for binocular or monocular visibility of 3D surface points. Given enough cells, the P system implementation speeds up the inner algorithm loop from O(nd) to O(n+d), where n is the width of a stereo image and d is the disparity range. The implementation gives also an insight into a more general SDPS that accounts for a possible multiplicity of solutions of the ill-posed problem of optimal stereo matching.

A GPU-Based Parallel Processing Method for Slope Analysis in Geographical Computation

Nowadays, geographical computation presents to be distributed, parallel, and diversification application trend. Influence of problem scale and response speed requirement received more attention. And high performance computational systems, such as “TianHe 1-A”, provides a new generation of hardware supports. In order to make full use of these high performance computational resources, appropriate and efficient parallel algorithms are needed. New parallel computing optimization technique of the geography is proposed in this paper by designing new parallel algorithms for slope analysis. We implement it based on CUDA. Experimental results with random DEM data in uniform distribution validate our methods.

ChemInform Abstract: Capturing Functional Motions of Membrane Channels and Transporters with Molecular Dynamics Simulation

AbstractReview: 179 refs.

MODELING AND SIMULATION OF MICRO-SCALE WIND FARMS USING HIGH PERFORMANCE COMPUTING

Onshore wind farms usually consist of numerous horizontal axis wind turbines closely placed in clusters, and they are often cited on complex terrain. This paper proposes a computational framework for modeling and simulation of wind flow over micro-scale (and early meso-scale) wind farms using distributed memory, massively parallel high performance computing platforms. The present framework uses the Reynolds Averaged Navier–Stokes (RANS) to model the wind flow over the wind farms, as the flow is considered to be fully turbulent, isothermal and incompressible. The wind turbines installed in the wind farm are modeled by the virtual blade model (VBM). This technique considers the presence of a wind turbine's rotor implicitly through momentum sources placed in an actuator disc, yielding indirectly a pressure jump across the disk, which varies with its radius and azimuth. The nonlinear, aerodynamic interaction between the rotor wakes with each other and with the terrain of the wind farm is simulated by coupling the VBM with the governing flow field equations. In this manner, an efficient parallel algorithm for implementing the VBM was developed and integrated with parallel computational fluid dynamics (CFD) core simulation engine. The accuracy and performance of the proposed framework were confirmed through several test cases carried out on the IBM Blue Gene ultra-scale supercomputer.