Single-processor Computer Research Articles

Application of the MR-CSI Method for Three-Dimensional Imaging of the Triaxial Induction Measurements

We present a nonlinear inversion method for 3-D imaging of the near-wellbore region employing multiarray triaxial induction data. The method used is the so-called multiplicative-regularized contrast source inversion method. There are two main features of this method. The first is that the method does not require the explicit solution of the full forward problem at each iteration step. The second is that the regularizer is included as a multiplicative constraint that allows the automatic and adaptive determination of the regularization parameter. These two features enable us to robustly solve a large-scale inverse problem using single-processor present-day computer power.

Toward “green” mechanical simulations in materials science

Because of the developpement of materials science, there is a need to reduce the computational complexity of mechanical models. This paper aims to show that the Hyper Reduction method enables to reduce computational resources used for numerical simulations. Large mechanical models involving distributed nonlinearities require parallel computers to solve the governing equations related to these models. The proposed Hyper Reduction of such models provides reduced governing equations that enable simulations on a single-processor computer. This is achieved by using a reduced-basis and a selection of equilibrium equations of the detailed model. The use of a single processor during less time enables to save an amazing amount of the electrical energy during the numerical simulation.

Sums and Densities of Fully Coupled Anharmonic Vibrational States: A Comparison of Three Practical Methods

Three practical methods for computing sums and densities of states of fully coupled anharmonic vibrations are compared. All three methods are based on the standard perturbation theory expansion for the vibrational energy. The accuracy of the perturbation theory expansion is tested by comparisons with computed eigenvalues and/or experimental vibrational constants taken from the literature for three- and four-atom molecules. For a number of examples, it is shown that the X(ij) terms in the perturbation theory expansion account for most of the anharmonicity, and the Y(ijk) terms also make a small contribution; contributions from the Z(ijkl) terms are insignificant. For molecules containing up to approximately 4 atoms, the sums and densities of states can be computed by using nested DO-loops, but this method becomes impractical for larger species. An efficient Monte Carlo method published previously is both accurate and practical for molecules containing 3-6 atoms but becomes too slow for larger species. The Wang-Landau algorithm is shown to be practical and reasonably accurate for molecules containing approximately 4 or more atoms, where the practical size limit (with a single computer processor) is currently on the order of perhaps 50 atoms. It is shown that the errors depend mostly on the average number of stochastic samples per energy bin. An automated version of the Wang-Landau algorithm is described. Also described are the effects of Fermi resonances and procedures for deperturbation of the anharmonicity coefficients. Computer codes based on all three algorithms are available from the authors and can also be downloaded freely from the Internet (http://aoss.engin.umich.edu/multiwell/).

Toward “green” mechanical simulations in materials science

Because of the developpement of materials science, there is a need to reduce the computational complexity of mechanical models. This paper aims to show that the Hyper Reduction method enables to reduce computational resources used for numerical simulations. Large mechanical models involving distributed nonlinearities require parallel computers to solve the governing equations related to these models. The proposed Hyper Reduction of such models provides reduced governing equations that enable simulations on a single-processor computer. This is achieved by using a reduced-basis and a selection of equilibrium equations of the detailed model. The use of a single processor during less time enables to save an amazing amount of the electrical energy during the numerical simulation.

A High Performance Multifrontal Code for Linear Solution of Structures Using Multi-Core Microprocessors

A multifrontal code is introduced for the efficient solution of the linear system of equations arising from the analysis of structures. The factorization phase is reduced into a series of interleaved element assembly and dense matrix operations for which the BLAS3 kernels are used. A similar approach is generalized for the forward and back substitution phases for the efficient solution of structures having multiple load conditions. The program performs all assembly and solution steps in parallel. Examples are presented which demonstrate the code's performance on single and dual core processor computers.

Development of the collocations and least squares method

We propose and implement new, more general versions of the method of collocations and least squares (the CLS method) and, for a system of linear algebraic equations, an orthogonal method for accelerating the convergence of an iterative solution. The use of the latter method and the proper choice of values of control parameters, based on the results of investigating the dependence of the properties of the CLS method on these parameters, as well as some other improvements of the CLS method suggested in this paper, allow one to solve numerically problems for Navier-Stokes equations in a reasonable time using a single-processor computer even for grids as large as 1280 × 1280. In this case, the total number of unknown variables is ∼ 25 · 106. The numerical results for the problem of the lid-driven cavity flow of a viscous fluid are in good agreement with known results of other authors, including those obtained by means of schemes of higher approximation order with a small artificial viscosity. This and some other facts prove that the new versions of the CLS method make it possible to obtain an approximate solution with high accuracy.

Real-Time Computerized Annotation of Pictures

Developing effective methods for automated annotation of digital pictures continues to challenge computer scientists. The capability of annotating pictures by computers can lead to breakthroughs in a wide range of applications, including Web image search, online picture-sharing communities, and scientific experiments. In this work, the authors developed new optimization and estimation techniques to address two fundamental problems in machine learning. These new techniques serve as the basis for the Automatic Linguistic Indexing of Pictures - Real Time (ALIPR) system of fully automatic and high speed annotation for online pictures. In particular, the D2-clustering method, in the same spirit as k-means for vectors, is developed to group objects represented by bags of weighted vectors. Moreover, a generalized mixture modeling technique (kernel smoothing as a special case) for non-vector data is developed using the novel concept of Hypothetical Local Mapping (HLM). ALIPR has been tested by thousands of pictures from an Internet photo-sharing site, unrelated to the source of those pictures used in the training process. Its performance has also been studied at an online demo site where arbitrary users provide pictures of their choices and indicate the correctness of each annotation word. The experimental results show that a single computer processor can suggest annotation terms in real-time and with good accuracy.

WinMechLab: A Windows-based Software Tool for Real-time Control of Mechatronic Systems

Real-time control of industrial processes has become an important issue in the recent years. Advances in hardware and software technologies allow the use of single-processor computers to perform real–time tasks. The paper first explains the basic elements of PC–based real–time control systems. Then, it describes the general operation of WinMechLab, a real–time single–processor platform for the simulation and control of mechanisms based on MS–Windows. Finally, we show the application of this software tool in a simple control task of a direct–drive didactic robot arm.

From the Screen of Stone

The topics included in this column are: a problem with using Internet Explorer 7 with the AP-S digital archive; first looks at the Lenovo ThinkPad T60 and the X60 Tablet notebook computers; a computer case; a major problem with the newest versions of Symantec's Norton Internet Security suites; why you may want to hang on to your old single-processor computer to run CEM codes; another external USB drive for backup; some e-mail formatting considerations; why a fire safe may not be a safe place for computer media; computer insurance; Microsoft Office file converters; on destroying data on a hard drive; and HDTV

TU‐FF‐A4‐04: Intensity Modulation Patterns for Regional Exposure Control with Multiple Angle Slot Scan Imaging: Simulated Annealing Optimization Technique Approach

Purpose: To study the feasibility of using simulated annealing algorithm to determine the intensity modulation patterns for regional exposure control with multiple angle slot scan imaging. Method and Materials: We acquired a digital image from the chest radiography and then processed with a 2‐D Gaussian filter as an exposure equalization mask. Slot scanning exposures at evenly spaced angle between 0 and 180 degrees were used to achieve the ideal exposure distribution. An optimization technique, simulated annealing, was used to search the best intensity modulation patterns. This method is based on the theory of statistical physics and uses Boltzmann probability distribution to locate the minimum energy state. An objective function was mathematically constructed and Metropolis scheme was incorporated into the numerical computation. Various tuning parameters such as the control temperature setting for the annealing schedule were explored and the best combination was empirically chosen. We also calculated the percent root mean square error to quantify the results. Results: A wide range of scanning angles was tested in the study. For 8 projection angles, it took 10 minutes to complete the intensity modulation patterns search in a single processor computer and the percent root mean square error was 12.0%. The percent root mean square error can be further reduced by adding the number of scanning angles. Conclusion: Our study indicated that simulated annealing technique has the potential to determine the optimized intensity modulation patterns. Current work is focused on the reduction of both the computing time and percent root mean square error. Other optimization techniques such as the conjugate gradient method and the applications of parallel computing methods to accelerate the search algorithms are to be investigated.Supported in part by a research grant EB00117 by the National Institute of Biomedical Imaging and Bioengineering.

NUMERICAL COMPARISON OF WANG–LANDAU SAMPLING AND PARALLEL TEMPERING FOR MET-ENKEPHALIN

We compare the efficiency of two prominent techniques for simulation of complex systems: parallel tempering and Wang–Landau sampling. We show that both methods are of comparable efficiency but are optimized for different platforms. Parallel tempering should be chosen on multi-processor system while Wang–Landau sampling is easier to implement on a single-processor computer.

Parallel Computing in Civil Engineering

This issue marks the beginning of a new initiative in parallel computing. I would like to start with the statement by Grace Hopper 1910–1992 , pioneer computer scientist and U.S. Navy admiral, “In pioneer days they used oxen for heavy pulling, and when one ox couldn’t budge a log, they didn’t try to grow a larger ox but used more oxen. We shouldn’t be trying for bigger computers, but for more systems of computers.” The essence of this message is that it is very difficult to build a single processor computer that can handle the processing needs of any application. In other words, no matter how powerful a single processor computer is, there are always applications that require the power of parallel processing. The problems that we face as civil engineers are rapidly changing with increasing emphasis on homeland security, civil infrastructure protection, sustainability, and smart designs. These changing focuses put a premium on large-scale computation and real-time processing for many applications. Enabling parallel computation for these applications is not trivial due to inherent complexities of emerging parallel computing environments, lack of easily usable software libraries for these environments, and the dynamic nature of data and computation needs for some real-time processing applications. Consumer needs and economic reality have changed the face of parallel computing dramatically over the last two decades with the infusion of a new generation of supercomputers with hierarchical memory subsystems, nonuniform processor configurations, and heterogeneous geographically distributed grid computing environments. Obtaining good performance on these environments for real-world applications is extremely challenging even with available middleware tools such as MPI, Globus, Matlab distributed computing toolbox, and parallel numerical software such as PETSc, SCALAPACK, DAKOTA, ParMETIS, and Aztec. Therefore, prospective users such as civil engineers are forced to develop their own algorithms and/or implementations for their application needs. Despite the fact that parallel computing has been an active research area for the last two decades and that the demand for parallel computing is increasing in civil engineering, the leading computing journal in civil engineering, JCCE, has seen very few papers in this area averaging about one article per year . This editorial’s purpose is to change this trend by issuing a call to the civil engineering community to support an initiative in parallel computing. I will next address the type of parallel computing articles we would like to publish in JCCE, i.e., what do we want in a parallel computing article for JCCE? Prospective authors may consider the following questions:

Development and Tests of a New Distributed-Memory MM5 Adjoint

Abstract Updated versions of the Tangent Linear Model (TLM) and adjoint of the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5) have been developed and are now available to the meteorological community. The previous version of the MM5 TLM and adjoint were designed for single-processor computer architectures, based on version 1 of MM5, and were hand coded, which made it difficult to maintain up-to-date versions of the TLM and the adjoint as MM5 evolved. The new TLM and adjoint are based on version 3 of MM5 and run efficiently on multiple-processor computers. The TLM and adjoint were developed with the aid of the Tangent Linear and Adjoint Model Compiler (TAMC) automatic code generator. While some manual intervention is still necessary, the use of the automatic code generator can significantly speed code development and lower code maintenance costs. The new TLM and adjoint contain most of the physics packages and observation operators that were available in the MM5 version 1 TLM and adjoint. The new adjoint has been combined with the MM5 version 3 nonlinear model and an updated minimization module in a four-dimensional variational data assimilation analysis configuration. Accuracy of the new TLM and adjoint has been verified by individual unit and system tests as well as comparisons with the adjoint from MM5 version 1. Timing tests showed substantial decreases in time to solution when increasing the number of processors devoted to the problem.

Scheduling flexible manufacturing systems using parallelization of multi-objective evolutionary algorithms

Solving multi-objective scientific and engineering problems is, generally, a very difficult goal. In these optimization problems, the objectives often conflict across a high-dimensional problem space and require extensive computational resources. In this paper, a migration model of parallelization is developed for a genetic algorithm (GA) based multi-objective evolutionary algorithm (MOEA). The MOEA generates a near-optimal schedule by simultaneously achieving two contradicting objectives of a flexible manufacturing system (FMS). The parallel implementation of the migration model showed a speedup in computation time and needed less objective function evaluations when compared to a single-population algorithm. So, even for a single-processor computer, implementing the parallel algorithm in a serial manner (pseudo-parallel) delivers better results. Two versions of the migration model are constructed and the performance of two parallel GAs is compared for their effectiveness in bringing genetic diversity and minimizing the total number of functional evaluations.

Simulation of the Big Elk Fire using coupled atmosphere - fire modeling

Models that simulate wildland fires span a vast range of complexity; the most physically complex present a difficult supercomputing challenge that cannot be solved fast enough to become a forecasting tool. Coupled atmosphere–fire model simulations of the Big Elk Fire, a wildfire that occurred in the Colorado Front Range during 2002, are used to explore whether some factors that make simulations more computationally demanding (such as coupling between the fire and the atmosphere and fine atmospheric model resolution) are needed to capture wildland fire parameters of interest such as fire perimeter growth. In addition to a Control simulation, other simulations remove the feedback to the atmospheric dynamics and use increasingly coarse atmospheric resolution, including some that can be computed in faster than real time on a single processor. These simulations show that, although the feedback between the fire and atmosphere must be included to capture accurately the shape of the fire, the simulations with relatively coarse atmospheric resolution (grid spacing 100–500 m) can qualitatively capture fire growth and behavior such as surface and crown fire spread and smoke transport. A comparison of the computational performance of the model configured at these different spatial resolutions shows that these can be performed faster than real time on a single computer processor. Thus, although this model still requires rigorous testing over a wide range of fire incidents, it is computationally possible to use models that can capture more complex fire behavior (such as rapid changes in intensity, large fire whirls, and interactions between fire, weather, and topography) than those used currently in the field and meet a faster-than-real-time operational constraint.

New Fast SCFT Algorithm Applied to Binary Diblock Copolymer/Homopolymer Blends

We present an efficient strategy for mapping out the classical phase behavior of block copolymer systems using self-consistent-field theory. With our new algorithm, the complete solution of a classical block copolymer phase can be evaluated typically in a fraction of a second on a single-processor computer, even for highly segregated melts. This is accomplished by implementing the standard unit-cell approximation for the cylindrical and spherical phases and solving the resulting equations using a Bessel function expansion. Here the method is used to investigate blends of AB diblock copolymer and A homopolymer, concentrating on the situation where the two molecules are of similar size.

An Efficient Parallel ADI Algorithm for Turbomachinery Flows

The paper describes a domain decomposition strategy which allows high efficiency parallel simulations of turbomachinery flows. The implicit ADI parallel solver is based on the scalar approximate factorisation. The Navier-Stokes (NS) and turbulence model equations are discretised by centred finite differences. The results prove that the parallel calculations with domain decomposition, in which each sub-domain explicitly communicates with the adjacent ones at the end of each implicit iteration, may suffer from a considerable deterioration of the convergence rate. A simple sub-iterative domain recoupling strategy allows recovering the convergence rate of a single processor computation. The strategy is carefully analysed and optimised in terms of inter-processor data communication rate and algorithm memory requirements. The span-wise domain decomposition is particularly suited for turbomachinery flows and is applied to a radial impeller and to an axial turbine stator and stage to prove the validity and the accuracy of the proposed approach. The results indicate that the parallel recoupled algorithm usually reach efficiencies of 0.8, with peaks over 0.9 with up to 16 processors, thereby allowing a considerable speed-up of design and verification calculations.

A real-time traffic simulation using a communication latency hiding parallelization

This paper implements and analyzes a highway traffic-flow simulation based on continuum modeling of traffic dynamics. A traffic-flow simulation was developed and mapped onto a parallel computer architecture. Two algorithms (the one-step and two-step algorithms) to solve the simulation equations were developed and implemented. Tests with real traffic data collected from the freeway network in the metropolitan area of Minneapolis, MN, were used to validate the accuracy and computation rate of the parallel simulation system (with 256 processors): The execution time for a 24-h traffic-flow simulation over a 15.5-mi freeway, which takes 65.7 min on a typical single-processor computer, took only 88.51 s on the nCUBE2 and only 2.39 s on the CRAY T3E. The two-step algorithm, whose goal is to trade off extra computation for fewer interprocessor communications, was shown to save significantly on the communication time on both parallel computers.

Use of randomization testing to detect multiple epistatic QTLs.

Here, we describe a randomization testing strategy for mapping interacting quantitative trait loci (QTLs). In a forward selection strategy, non-interacting QTLs and simultaneously mapped interacting QTL pairs are added to a total genetic model. Simultaneous mapping of epistatic QTLs increases the power of the mapping strategy by allowing detection of interacting QTL pairs where none of the QTL can be detected by their marginal additive and dominance effects. Randomization testing is used to derive empirical significance thresholds for every model selection step in the procedure. A simulation study was used to evaluate the statistical properties of the proposed randomization tests and for which types of epistasis simultaneous mapping of epistatic QTLs adds power. Least squares regression was used for QTL parameter estimation but any other QTL mapping method can be used. A genetic algorithm was used to search for interacting QTL pairs, which makes the proposed strategy feasible for single processor computers. We believe that this method will facilitate the evaluation of the importance at epistatic interaction among QTLs controlling multifactorial traits and disorders.

Parallel processing for Rietveld refinement

A method for speeding Rietveld refinements using parallel computing is presented. The method can be applied to most, if not all, Rietveld refinement programs. An example implementation for the Los AlamosGeneral Structure Analysis System(GSAS) package is described. Using a cluster with seven processors, least-squares refinements are completed 2.5 to 5 times faster than on an equivalent single-processor computer.