Multiprocessor Computer Research Articles

An extended generalized hypercube as a fault-tolerant system area network for multiprocessor systems

This paper suggests a novel structure of a system area network for high-performance multiprocessor computers. The authors consider a system area network in the form of an extended generalized hypercube, where complete graph topology links in rows are replaced by quasicomplete graph topology links with an appreciably smaller number of edges.

Performance Analysis of CPU-GPU Cluster Architectures

High performance computing (HPC) encompasses advanced computation over parallel processing, enabling faster execution of highly compute intensive tasks such as climate research, molecular modeling, physical simulations, cryptanalysis, geophysical modeling, automotive and aerospace design, financial modeling, data mining and more. High performance simulations require the most efficient compute platforms. The execution time of a given simulation depends upon many factors, such as the number of CPU/GPU cores and their utilization factor and the interconnect performance, efficiency, and scalability. CPU and GPU clusters are one of the most progressive branches in a field of parallel computing and data processing nowadays. GPUs have become increasingly common in supercomputing, serving as accelerators or "co-processors" in every node CPU-GPU to help CPUs get work done faster. In this paper I use the Multiclass Closed Product-Form Queueing Network (MCPFQN) and Mean Value Analysis (MVA) to analyze effects of the CPU-GPU cluster interconnect on the performance of computer systems.

Methods and Algorithms for the Solution of Inverse Problems of Modelling and Control in Reservoir Engineering

Statement and solution of the problems of modelling and control of hydrodynamic parameters in reservoir engineering are usually associated with insufficiency and/or inaccuracy of the original input data (initial and boundary conditions) and the parameters of filtration equations in porous medium of reservoirs. In such conditions it is necessary to solve the identification problem (machinery monitoring) of the rough specified initial data (initial and boundary conditions), as well as parameters (permeability coefficients) of hydrodynamic models of reservoirs by historical data. This means that the traditional hydrodinamic models, which are formulated as direct initial boundary value problems for the equations of fluid filtration are improperly posed, i.e. small changes in the input data can lead to an arbitrarily large changes in the output. Therefore the models for the control of the field development are formulated as inverse problems and the iterative regularization methods are offered for their solutions. Problem of high dimensionality naturally arises under these statements of the problems of modelling and control, to solve which the multilevel parallel computing technologies based on the hierarchical decomposition combined with the multigrid versions of the methods of splitting by physical processes, spatial and temporal coordinates are offered. Dimension of the problem is even greater increases under development of offshore fields and coastal areas (e.g. in the Arctic) in real-time when it is required to create the integrated simulation models of the complex technological system (reservoir - well - oil and gas collecting network) as a whole.To solve the problems of identification and adaptation of initial and boundary conditions and filtration parameters the iterative regularization methods for direct and inverse problems for the original model equations by historical data are offered. To create the integrated model of the complex technological system as a whole the innovative parallel computing technologies with optimal hierarchical (multilevel) embedding of parallel algorithms into the architecture of multiprocessor computer systems using MPI/OpenMP or OpenCL/CUDA are offered.

SHELXT - integrated space-group and crystal-structure determination.

The new computer program SHELXT employs a novel dual-space algorithm to solve the phase problem for single-crystal reflection data expanded to the space group P1. Missing data are taken into account and the resolution extended if necessary. All space groups in the specified Laue group are tested to find which are consistent with the P1 phases. After applying the resulting origin shifts and space-group symmetry, the solutions are subject to further dual-space recycling followed by a peak search and summation of the electron density around each peak. Elements are assigned to give the best fit to the integrated peak densities and if necessary additional elements are considered. An isotropic refinement is followed for non-centrosymmetric space groups by the calculation of a Flack parameter and, if appropriate, inversion of the structure. The structure is assembled to maximize its connectivity and centred optimally in the unit cell. SHELXT has already solved many thousand structures with a high success rate, and is optimized for multiprocessor computers. It is, however, unsuitable for severely disordered and twinned structures because it is based on the assumption that the structure consists of atoms.

Effect of bacteria density and accumulated inert solids on the effluent pollutant concentrations predicted by the constructed wetlands model BIO_PORE

Constructed wetlands are a widely adopted technology for the treatment of wastewater in small communities. The understanding of their internal functioning has increased at an unprecedented pace over recent years, in part thanks to the use of mathematical models. BIO_PORE model is one of the most recent models developed for constructed wetlands. This model was built in the COMSOL Multiphysics™ software and implements the biokinetic expressions of Constructed Wetlands Model 1 (CWM1) to describe the fate and transport of organic matter, nitrogen and sulphur in horizontal subsurface-flow constructed wetlands. In previous studies, CWM1 was extended with the inclusion of two empirical parameters (Mbio_max and Mcap) that proved to be essential to provide realistic bacteria growth rates and dynamics. The aim of the current work was to determine the effect of these two parameters on the effluent pollutant concentrations predicted by the model. To that end, nine simulations, each with a different Mbio_max-Mcap pair, were launched on a high-end multi-processor computer and the effluent COD and ammonia nitrogen concentrations obtained on each simulation were qualitatively compared among them. Prior to this study, a finite element mesh optimization procedure was carried out to reduce computational cost. Results of the mesh optimization procedure indicated that among the 5 tested meshes of different element size, the mesh utilized for this model in previous studies represented a fair compromise between output accuracy and computation time. Results of the sensitivity analysis showed that the value of Mcap has a dramatic effect on the simulated effluent concentrations of COD and ammonia nitrogen, which clearly decreased for increasing values of this parameter. On the other hand, the model output was also sensitive to the values of Mbio_max, but its effects were less important and no clear relation could be established between its value and the simulated effluent concentration of COD and ammonia nitrogen.

Parallel domain decomposition method with non-blocking communication for flow through porous media

This paper introduces a domain decomposition method for numerically solving the Stokes equation for very large, complex geometries. Examples arise from realistic porous media. The computational method is based on the SIMPLE (Semi-Implicit Method for Pressure Linked Equations) algorithm which uses a finite-differences approach for discretizing the underlying equations. It achieves comparable speed and efficiency as lattice Boltzmann methods. The domain decomposition method splits a large three-dimensional region into slices that can be processed in parallel on multi-processor computation environments with only minimal communication between the computation nodes. With this method, the flow through a porous medium with grid sizes up to 20483 voxel has been calculated.

A new parallel lattice reduction algorithm for BKZ reduced bases

In order to implement the original BKZ algorithm in parallel, we describe it in terms of parallelism and give its parallel implementation scheme. Then we analyze the efficiency of algorithm’s parallel implementation and show that the speedup factor of BKZ algorithm in parallel is extremely low. Therefore we present a new parallel lattice reduction algorithm suitable for multiprocessor computer architecture. The new algorithm can obtain a BKZ reduced basis and the parallel speedup is effective. Also with the practical results, although the computational complexity increases compared with the original BKZ algorithm, we still indicate that the new algorithm performs well in parallel and the time cost in parallel is less. At the same time, we show that the length of the shortest vector is smaller.

Sufficient conditions for the determination and use of data in the same granular parallel computation process

Parallel algorithms for distributed memory computers should be granular, in which case the set of algorithmic operations is split into sets known as computation grains, or tiles. Conditions are proposed and proved under which data is used in the same granular computation process where it was determined. These conditions can be used to estimate the number of communication operations in alternative versions of parallel algorithms.

Using a sparse promoting method in linear programming approximations to schedule parallel jobs

SummaryIn this paper, we tackle the well‐known problem of scheduling a collection of parallel jobs on a set of processors either in a cluster or in a multiprocessor computer. For the makespan objective, that is, the completion time of the last job, this problem has been shown to be NP‐hard, and several heuristics have already been proposed to minimize the execution time. In this paper, we consider both rigid and moldable jobs. Our main contribution is the introduction of a new approach to the scheduling problem, based on the recent discoveries in the field of compressed sensing. In the proposed approach, all possible positions and shapes of the jobs are encoded into a matrix, and the scheduling is performed by selecting the best columns under natural constraints. Thus, the solution to the new scheduling formulation is naturally sparse, and we may use appropriate relaxations to achieve the optimization task in the quickest possible way. Among many possible relaxation strategies, we choose to minimize the ℓp‐quasi‐norm for p∈(0,1). Minimization of the ℓp‐quasi‐norm is implemented via a successive linear programming approximation heuristic. We propose several new algorithms based on this approach, and we assess their efficiency through simulations. The experiments show that the scheme outperforms the classic Largest Task First list based algorithm for scheduling small to medium instances but needs improvements to compete on larger numbers of jobs. Copyright © 2014 John Wiley & Sons, Ltd.

Automated Brain Tumor Detection in Medical Brain Images and Clinical Parameters using Data Mining Techniques: A Review

mining is a growing field of research that intersects with many other fields such as Artificial Intelligent, Statistics, Visualization, Parallel Computing and Image Processing. Data mining techniques are good for Brain MRI image classification that can diagnose brain tumor and other diseases. In this paper we present an overview of the current research being carried out using the data mining techniques for the diagnosis of brain tumor. The goal of this study is to identify the most well performing data mining algorithms used on medical brain MRI and Clinical parameters. The following algorithms have been identified: Decision Trees, Support Vector Machine, Artificial Neural Networks and their Multilayer Perceptron model, and Fuzzy C-Means. Analyses show that it is very difficult to name a single data mining algorithm as the most suitable for the brain tumor detection or classification. At times some algorithms perform better than others, but there are cases when the properties of some of the above mentioned algorithm are combined together, they provide effective result. This paper also provides a critical evaluation of the literature reviewed, which reveals new facets of.

Parallelization of nonuniform loops in supercomputers with distributed memory

A template algorithm for parallel execution of independent iterations of the repetitive loop on a multiprocessor computer with distributed memory is constructed. Regardless of the number of processors, the algorithm must provide efficient utilization of computing capacity under essentially different complexities of iterations and/or performance of processors. The interprocessor data communication and control of par� allel computations are assumed to be implemented using a standard messagepassing interface (MPI), which is widely used in such systems. Existing methods for the loop parallelization are analyzed and the correspond� ing efficiencies are empirically estimated for various models of iteration nonuniformity.

О некоторых вариантах методов декомпозиции областей

The paper considers the algorithms for solving large sparse SLAEs arising from grid approximations of boundary value problems. The SLAEs and algorithms are not limited in a sense of number of unknowns, computational nodes, processors and/or cores. This problem is reduced to a distributed variant of algebraic 3D-domain decomposition, in which no excessive load of the root process is present, i.e. all MPI-processes, each of which corresponds to its own subdomain, are almost equal. The computational process consists of two main stages. The first stage is the automatic decomposition, based on the analysis of the matrix portrait and the formation of large-block representation of the original SLAE. The second stage implements a Krylov subspace iterative process with FGMRes (flexible generalized minimal residual method) using either exact or approximate inverse of diagonal blocks as a preconditioner. The methods described are implemented as a part of Krylov, a library of algebraic solvers. The paper presents some features of current parallel implementation and estimates of resource usage. Efficiency of the developed algorithms is illustrated by solving several typical model problems with different parameters and in different configurations of multiprocessor computer systems.

Emulation and Analytical Model of PIM Supplemental Computing Element via HDL

Abstract This article presents and describes the simulation and emulation of single processing core, designed for new heterogeneous multiprocessor computer architectures with supplemental computing elements in the operating memory. We study the emulated results over the simulated model to confirm it’s usability in the overall simulation of complete system. The simulated models are developed based on existing data, comparing time markers, measuring the output performance against the results from the emulation. Using additional computing elements in the memory will allows us to improve the current limitations of the conventional architectures.

Imitation Monte Carlo methods for problems of the Boltzmann equation with small Knudsen numbers, parallelizing algorithms with splitting

The talk examines a system of pairwise interaction particles, which models a rarefied gas in accordance with the nonlinear Boltzmann equation, the master equations of Markov evolution of this system and corresponding numerical Monte Carlo methods. Selection of some optimal method for simulation of rarefied gas dynamics depends on the spatial size of the gas flow domain. For problems with the Knudsen number Kn of order unity "imitation", or "continuous time", Monte Carlo methods ([2]) are quite adequate and competitive. However if Kn ≤ 0.1 (the large sizes), excessive punctuality, namely, the need to see all the pairs of particles in the latter, leads to a significant increase in computational cost(complexity). We are interested in to construct the optimal methods for Boltzmann equation problems with large enough spatial sizes of the flow. Speaking of the optimal, we mean that we are talking about algorithms for parallel computation to be implemented on high-performance multi-processor computers. The characteristic property of large systems is the weak dependence of sub-parts of each other at a sufficiently small time intervals. This property is taken into account in the approximate methods using various splittings of operator of corresponding master equations. In the paper, we develop the approximate method based on the splitting of the operator of master equations system "over groups of particles" ([7]). The essence of the method is that the system of particles is divided into spatial subparts which are modeled independently for small intervals of time, using the precise"imitation" method. The type of splitting used is different from other well-known type "over collisions and displacements", which is an attribute of the known Direct simulation Monte Carlo methods. The second attribute of the last ones is the grid of the "interaction cells", which is completely absent in the imitation methods. The main of talk is parallelization of the imitation algorithms with splitting using the MPI library. New constructed algorithms are applied to solve the problems: on propagation of the temperature discontinuity and on plane Poiseuille flow in the field of external forces. In particular, on the basis of numerical solutions, comparative estimates of the computational cost are given for all algorithms under consideration.

A Sparse Self-Consistent Field Algorithm and Its Parallel Implementation: Application to Density-Functional-Based Tight Binding.

We present an algorithm and its parallel implementation for solving a self-consistent problem as encountered in Hartree-Fock or density functional theory. The algorithm takes advantage of the sparsity of matrices through the use of local molecular orbitals. The implementation allows one to exploit efficiently modern symmetric multiprocessing (SMP) computer architectures. As a first application, the algorithm is used within the density-functional-based tight binding method, for which most of the computational time is spent in the linear algebra routines (diagonalization of the Fock/Kohn-Sham matrix). We show that with this algorithm (i) single point calculations on very large systems (millions of atoms) can be performed on large SMP machines, (ii) calculations involving intermediate size systems (1000-100 000 atoms) are also strongly accelerated and can run efficiently on standard servers, and (iii) the error on the total energy due to the use of a cutoff in the molecular orbital coefficients can be controlled such that it remains smaller than the SCF convergence criterion.

Software Demodulation Method of PCM/FM Signal Based on the Maximum Likelihood Estimation

Describes the establishment of maximum likelihood estimation model and victor than the realization of the search path. At the same time according to the data block transmission mode in the software and the characteristics of victor than path search, research a kind of software demodulation is implemented based on data redundancy, make full use of the parallel processing computer system performance, improve the operation efficiency of the algorithm. And the algorithm error rate as an index, compares the software likelihood estimation algorithm with the traditional frequency demodulation performance, results show that the BER 10-4 when algorithm can get the demodulation gain of about 3 dB.

Domain Decomposition Algorithms for Edge Element Based Parabolic Type Problems

Today, a huge amount of interest in new actuators and a correspondingly huge pressure on companies and research centers to develop the most efficient, cost-effective electric design of the actuators for electric vehicles. However, the design and analysis of actuators is very complex task, thus need the help of multiprocessor computers to run efficient computations. This work analyzes the use of two non-overlapping domain decomposition methods (DDMs) in order to improve the calculation behaviour of finite element method (FEM) with edge element approximation. In this case, the DDMs under investigation are the Schur complement method and the Lagrange multiplier based Finite Element Tearing and Interconnecting (FETI) method with their algorithms. The performance of these methods and solvers have been investigated in detail for two-dimensional parabolic type problems as case studies.

From medical images to flow computations without user-generated meshes.

Biomedical flow computations in patient-specific geometries require integrating image acquisition and processing with fluid flow solvers. Typically, image-based modeling processes involve several steps, such as image segmentation, surface mesh generation, volumetric flow mesh generation, and finally, computational simulation. These steps are performed separately, often using separate pieces of software, and each step requires considerable expertise and investment of time on the part of the user. In this paper, an alternative framework is presented in which the entire image-based modeling process is performed on a Cartesian domain where the image is embedded within the domain as an implicit surface. Thus, the framework circumvents the need for generating surface meshes to fit complex geometries and subsequent creation of body-fitted flow meshes. Cartesian mesh pruning, local mesh refinement, and massive parallelization provide computational efficiency; the image-to-computation techniques adopted are chosen to be suitable for distributed memory architectures. The complete framework is demonstrated with flow calculations computed in two 3D image reconstructions of geometrically dissimilar intracranial aneurysms. The flow calculations are performed on multiprocessor computer architectures and are compared against calculations performed with a standard multistep route.

Parallel Algorithms in a Physically-Based Hydrologic Soil Model

Enormous river basin information has been collected by for high resolution of the physically-based distributed hydrological model, while the scales of computational domain are often restricted by the intensive calculation. The parallel algorithms of physically-based regional hydrologic soil model (P-RHSM) in OpenGeoSys are brought forward in this paper. The object-oriented C++ programming in the model approaches the two objectives: one is data structure organizations including the spatial discretization of the computation domain with nodes density arrangement and solutions communication of temporal variables, the other is computational efficiency by parallel computing. The Message passing interface (MPI) communication is implemented in P-RHSM which is carried out on Linux supported symmetric multiprocessing computers. P-RHSM is applied to Heihe basin in the northwest China. The high-resolution results of temporal soil moisture content distribution, front movement and parallel performance is presented.

A numerical model for the transient analysis of offshore foundations under cyclic loading

A comprehensive numerical model for the analysis of offshore foundations under a general transient loading is presented here. The theoretical basis of the model lies on the Swansea formulation of Biot’s equations of dynamic poroelasticity combined with a constitutive model that reproduces key aspects of cyclic soil behaviour in the frame of the theory of generalised plasticity. On the practical side, the adoption of appropriate finite element formulations may prevent the appearance of spurious numerical instabilities of the pore pressure field. In this respect, the use of a coupled enhanced-strain element is here proposed. On the other hand, the practicality of the presented model depends ultimately on its computational efficiency. Some practical recommendations concerning the solution strategies, the matrix storage/handling procedures and the parallel multi-processor computation are here provided. Finally, the performance of the model with a benchmark study case and its practical application to analyse the soil–structure interaction of an offshore monopile under a realistic transient storm loading are discussed.