http://www.computer.org/cise Physics, medicine, astronomy -- these and other hard sciences share a common need for efficient algorithms, system software, and computer architecture to address large computational problems. And yet, useful advances in computational techniques that could benefit many researchers are rarely shared. To meet that need, Computing in Science & Engineering presents scientific and computational contributions in a clear and accessible format. en-us Wed, 4 Jan 2012 11:00:01 GMT http://csdl.computer.org/common/images/logos/cise.gif List of recently published journal articles http://www.computer.org/cise http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.118 This paper presents a GPU-based high performance computing method for the protein-DNA docking problem. The protein-DNA docking algorithm in consideration integrates Monte-Carlo simulation and simulated annealing method. We investigate key factors that affect performance on a GPU cluster. These include the arrangement of the computation kernel and the balance between memory coalescing and path divergence. We also compared different Monte-Carlo schemes with respects to computation cost and prediction accuracy. With optimizations of these design options, our algorithm achieved 10.4 TFLOPS of sustained performance using 128 Tesla M2070 GPU cards, which represents 3.6x speed up over a traditional cluster with 1000 CPU cores. We show that such improved computation capability accelerates the computation of the protein-DNA docking problem, which leads to improved docking accuracy. To the best of our knowledge, this is the first ad hoc effort of applying GPU or GPU clusters to the protein-DNA docking problem. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.118 http://doi.ieeecomputersociety.org/10.1109/MCSE.2012.1 Algorithms designed to efficiently solve this classical problem of physics fit very well on GPU hardware, and exhibit excellent scalability on many GPUs. Their computational intensity makes them a promising approach for many other applications amenable to an N-body formulation. Adding features such as auto-tuning makes multipole-type algorithms ideal for heterogeneous computing environments. http://doi.ieeecomputersociety.org/10.1109/MCSE.2012.1 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.119 GPUs are low cost highly parallel devices that are now not only used for graphics but also for numerical simulation. We present three applications of a computer system with multiple GPUs to the domain of theoretical neuroscience. The first application is to a continuous model of the primary visual area, the second to the simulation of a stochastic neural network, and the third to the computation of the probability distribution on the possible states of a network. In all three cases we show that the speed-up obtained by the use of GPUs has considerably helped answering a scientific or technological question. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.119 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.117 Unsteady computational fluid dynamics simulations of turbulence are performed using up to 64 graphics processors. The results from two GPU clusters and a CPU cluster are compared. A second-order staggered-mesh spatial discretization is coupled with a low storage three-step Runge-Kutta time advancement and pressure projection at each substep. The pressure Poisson equation dominates the solution time and is solved with the preconditioned Conjugate Gradient method. The CFD algorithm is optimized to use the fast shared-memory on the GPUs and to use communication/computation overlapping. Detailed timings reveal that the internal calculations now occur so efficiently that the operations related to communication are the scaling bottleneck at all but the very largest problem sizes that can fit on the hardware. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.117 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.115 Testing a scientific framework is a challenging task, since, among other typical challenges in testing scientific software, one has to find a way to deal with the large variability in a framework. Our approach is to apply software product line engineering to handle the frameworks variability. We use variability modeling to support the selection of test applications and test cases. In this article we introduce the regression test environment we developed for DUNE, a complex scientific framework. The test environment consists of system tests that are constructed to take the huge variability of possible uses for the framework into account. We use system testing, since it is the only testing level where the interaction between the mathematical model, the numerical model and its implementation can be thoroughly tested. The tests also support algorithm verification and scientific validation. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.115 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.116 Modern scientific enterprises are inherently knowledge-intensive. Scientific studies in domains such as geosciences, climate modeling and biology require the acquisition and manipulation of experimental and field data to create inputs for large-scale computational simulations. The results of this process must be managed to provide justifications for regulatory decisions and scientific publications. In this paper we introduce our Velo framework, a reusable, domain independent knowledge management infrastructure for modeling and simulation. Velo leverages, integrates and extends Web based open source collaborative and data management technologies to create a scalable, flexible platform that can be tailored to specific scientific domains. We describe the architecture of Velo for managing and associating the various types of data that are used and created in modeling and simulation projects, as well as the framework for integrating domain-specific tools. To demonstrate realizations of Velo, we describe examples from two deployed sites for carbon sequestration and climate modeling. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.116 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.114 The visualization and exploration of large volumetric medical datasets from high resolution CT and MRI medical images are slow in rendering speed due the amount data to be processed. We present distance-based methods to address the issues based on GPU volume ray-casting and the idea of focus and context, providing a simultaneous visualization of different rendering styles for the focus region defined by a superquadric or a plane and its context. Experiments on large volumetric medical datasets show that our methods are efficient enough for interactive visualization and effective in directly guiding the user to region of interest (ROI). http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.114 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.113 The development of scientific software has similarities with processes that follow the software engineering Agile Mani-festo: responsiveness to change and collaboration are of utmost importance. But how well do current scientific software development processes match the practices found in agile development methods, and what are the effects of using agile practices in such processes? To find out, the authors conducted a literature review of past pro-jects and a multiple case study of three ongoing scientific software development projects. The results indicate that (1) no strong evidence of the general use of agile practices in projects developing scientific software is presently avail-able and (2) where agile practices are used this happens selectively. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.113 http://opac.ieeecomputersociety.org/opac?year=2011&volume=13&issue=06&acronym=cise Computing in Science and Engineering http://www.computer.org/portal/site/cise/ http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.96 We present a geodesic distance-based approach to synthesize natural facial animations using radial basis function (RBF). The geodesic distance calculation and the geodesic distance-based RBF interpolation. We first present an approximate method to calculate the geodesic distance according to three rules. For a mesh surface, the Euclidean distance may be quite different from the distance along the mesh surface, and this will be more obvious for a facial model with holes. The geodesic distance provides an approximate description for the distance along mesh surfaces, which is important for expression synthesis. Then, we present a new scheme to synthesize natural facial expressions using the geodesic distance instead of the Euclidean distance in the RBF interpolation. Experimental results show that our method can synthesize more natural expressions than the Euclidean distance-based RBF interpolation method and the “mesh distance”-based method, which is especially obvious in the regions around the mouth and the eyes. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.96 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.95 Recently a novel heuristic search algorithm, called Gravitational Search Algorithm (GSA), which is based on the law of gravity and mass interactions, has been proposed.. Although GSA has high performance in solving various optimization problems, it has some time consuming computations for calculation of the total force on each mass which makes the speed of optimization low. In this paper we introduce a new approach, which improves GSA's speed considerably. Our approach is based on the multi-agent systems where multiple agents are the mechanism used to express the parallelism. In multi-agent based GSA, complex problems are decomposed into smaller and simpler components that are handled by different agents in the system. Our experimental results show our multi-agent based GSA approach provides a high performance and high speed optimization methodology that can help scientists in a variety of science and engineering computations. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.95 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.94 We describe intriguing analogies between our model of plastic deformation in crystals and turbulence in liquids. The study of our model provides remarkable explanations of known experiments and predicts fractal dislocation pattern formation. The challenges that we encounter resemble the ones in turbulence, and we vividly exemplify them with a comparison to the Rayleigh-Taylor instability: they both have robust self-similar features with structures appearing on all scales, but they lack spatio-temporal convergence to a unique solution. Using this analogy, we show that tools that are successful for the study of turbulence can be utilized for dislocation dynamics. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.94 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.92 Computers are now extensively used as important tools throughout science, engineering, and medicine. Advances in computational geometric modeling, imaging, and simulation allow researchers to build and test models of increasing complexity and thus to generate unprecedented amounts of data. As noted in the NIH-NSF Visualization Research Challenges report [1], to effectively understand and make use of the vast amounts of data being produced is one of the greatest scientific challenges of the 21st Century. Visualization, namely helping researchers explore measured or simulated data to gain insight into structures and relationships within the data, will be critical in achieving this goal and is fundamental to understanding models of complex phenomena. In this paper, I highlight recent research in visualization as applied to biomedical applications, focusing on the work by researchers at the Scientific Computing and Imaging (SCI) Institute and their biomedical collaborators [2]. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.92 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.76 The European Theoretical Spectroscopy Facility (ETSF) is a distributed knowledge network that provides experimentalists and other researchers with access to state-of-the-art computer simulations for electronic excited states in matter. Its focus is on the fundamental knowledge of matter at the quantum-mechanical level and the transfer of this understanding to the future design of technologies in areas such as photovoltaics, light emitting diodes, optical data storage, and nanoelectronics. To bridge the gap between simulation and experiment, the ETSF is structured like an experimental light source facility, using the same nomenclature and processes. This article describes the operational and software aspects of the ETSF that enable it to interface efficiently with experimental research, as well as providing examples of ETSF research highlights and areas of applications. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.76 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.75 Research computing in radiation therapy is undergoing significant transition, from relatively small scale computing, to computing with large software systems. This evolution is driven by introduction of accelerator mounted imaging devices which collect multiple patient images at every treatment fraction, creating research datasets of hundreds of images per patient. Upgrades to software infrastructure can be very difficult to manage in an academic environment due to inadequate experience in large scale software development. This article describes our experience building a fairly large software system within a radiation oncology department, and emphasizes software architecture that promotes successful outcome.. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.75 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.56 We show a modelling method, based on frequency-domain system identification techniques combined with an optimal model-order selection for the efficient simulation of microwave radiometers. We model a number of different complex functions working with broadband signals with very different time scales. Taking into account the large number of different simulations required to design and test such a system, the selection of the optimal model-order is critical. Model parameters are signal independent and can be employed, in both frequency-domain simulations and in time-domain simulations. We illustrate the technique by its application in the modelling and simulation of a broadband microwave receiver, which is part of the 30-GHz radiometer of the Q U I Joint Tenerife (QUIJOTE) Cosmic Microwave Background (CMB) experiment. This system is a challenging case for conventional system simulation procedures because it samples the extremely low power broadband (10GHz) CMB signals over a long time. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.56 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.55 Dimensionality reduction, which consists in finding meaningful low dimensional structures hidden in high-dimensional observations, is a frequent problem for all scientists and engineers working with large volumes of data. The human brain confronts the same problem in everyday visual perception, extracting from its 1 000 000 optic nerve fibers a manageably small number of relevant features. We here present an approach to globally compute a descriptor of each image in a dataset. The adaptability of our approach allows one to combine it with existing dimensionality reduction methods and to benefit of its properties such as nonlinearity, global optimality, or convergence. The consequent bi-level procedure is capable of discovering the underlying global geometry of a complex natural observations dataset, such as human handwriting or faces under different viewing position, with a higher precision degree than existing methods. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.55 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.54 Density functional theory has become the dominant approach for simulating material properties in systems ranging from nanoscale electronics to crystals deep in the Earth’s core. However, the basis set used to determine the electron density in a particular density functional approach can have an important and subtle impact on our insight into a material. As I will demonstrate, the basis set chosen to describe graphene can affect interpretation of results from common characterization techniques and also electronic transport predictions for graphene devices. It is an issue that unfortunately is not emphasized enough in computational materials science and something that should be considered when choosing your weapon of choice for density functional calculations on a new material or nanostructure. For readers from other disciplines, it may also serve to underscore how the character of a basis set used to describe a function can impact our understanding. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.54 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.40 We describe a Monte Carlo based approach for simulating the kinetics on nanostructured catalytic and electrocatalytic systems relevant for industrial applications. A method to apply the Dynamic Monte Carlo (DMC) method to realistic catalyst geometries is presented. We also discuss the computational issues involved and present the application of the approach to an industrially relevant system - fuel cells. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.40 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.21 The reproducibility of research has become an intensively discussed topic in recent years. There are many good reasons why one would like scientific results to be reproducible, and a substantial amount of work has been done in order to set up a general framework that allows to achieve this goal. However, in modern computational simulation we can also observe a tendency towards the use of high performance computing systems. These systems usually contain massively parallel computing hardware and corresponding software specially tuned in order to minimize the time required for the computation. It is the goal of this paper to point out that such a solely speed-oriented approach often conflicts with the strive for reproducibility. Thus we obtain simulations that behave reproducible only in a limited way. We present the technical background of this phenomenon. It turns out that it will be very difficult to overcome these problems. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.21 http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.17 Legacy scientific codes are often re-purposed to fit adaptive needs, such as to dynamically alter parameters to improve convergence behavior, or to switch algorithms at runtime for greater accuracy of modeling. Given a legacy scientific code, how can we make it adaptive without making changes to the original source program(s)? We present an approach—Adaptive Code Collage (ACC)—to achieve this goal using function call interception in a language-neutral way at link time. ACC transparently ‘catches’ function calls and redirects them so that an existing program can be made adaptive without causing a significant performance overhead. We demonstrate the application of ACC to designing adaptive SOR algorithms for solving linear systems and to improving the performance, stability, and accuracy of GenIDLEST, a large parallel computational fluid dynamics code. http://doi.ieeecomputersociety.org/10.1109/MCSE.2011.17 http://doi.ieeecomputersociety.org/10.1109/MCSE.2010.144 Data assimilation aims to decrease errors in initial conditions of numerical weather prediction models, which are a primary source of uncertainty in hurricane prediction. This study examines the performance of three advanced techniques that assimilate inner-core, high-resolution Doppler radar observations for cloud-resolving hurricane initialization and forecasting for Hurricane Katrina. http://doi.ieeecomputersociety.org/10.1109/MCSE.2010.144 http://doi.ieeecomputersociety.org/10.1109/MCSE.2010.122 Continuum quantum Monte Carlo (QMC) has proved to be an invaluable tool for predicting the properties of matter from fundamental principles. By solving the many-body Schr¨odinger equation through a stochastic projection, it achieves greater accuracy than mean-field methods and much better scalability than quantum chemical methods, enabling scientific discovery across a broad spectrum of disciplines. The multiple forms of parallelism afforded by QMC algorithms make them ideal candidates for acceleration in the many-core paradigm. We present the results of our effort to port the QMCPACK simulation code to the NVIDIA CUDA GPU platform. We restructure the CPU algorithms to express additional parallelism, minimize GPU-CPU communication, and efficiently utilize the GPU memory hierarchy. Using mixed precision on GT200 GPUs and MPI for intercommunication and load balancing, we observe typical full-application speedups of approximately 10x to 15x relative to quad-core Xeon CPUs alone, while reproducing the double-precision CPU results within statistical error. http://doi.ieeecomputersociety.org/10.1109/MCSE.2010.122