Single-processor Workstation Research Articles

Parallel variable-resolution bathymetric estimation with static load balancing

A method for partitioning a large computation task (direct, variable resolution bathymetric grid construction from raw observations) into thread-parallel code is described. Based on the data density estimated for the first pass of the chrt algorithm, this algorithm statically partitions the estimation task into spatially distinct blocks of approximately equal total data observation count so that each can be executed in parallel and be expected to complete approximately concurrently. No communication between blocks or further load balancing is therefore required. A branch-and-bound algorithm is used to control the complexity of the partitioning task, but the computation time increases significantly as more partitions are required, leading to a degree of diminishing returns for allocating further computational resources and suggesting alternative approaches for high thread-count systems. Speed-up of the algorithm over a pair of test datasets (using real-world hydrographic survey data) shows that the performance consistently improves with the number of computational tasks assigned, initially (super-) linearly, although ultimately sub-linearly as other resource sharing limitations take over. An overall speedup of 4.1 times is demonstrated with a quad-core single-processor workstation.

Implementation of a pre-calculated database approach for scatter correction in SPECT

In single photon emission computed tomography (SPECT), attenuation and scatter introduce important artefacts in the reconstructed images leading to mis diagnosis for patient’s follow-up. Furthermore, by using Monte Carlo simulation (MCS), physical effects undergone by photons during the SPECT exam can be precisely modeled and accounted for during iterative reconstruction, which improves the quality of the image. However, MCS are time consuming and therefore inappropriate for the rate of daily exams performed in clinical routine. Our work is based on the assumption that patients are composed of identical biological tissues and that photon propagation in an element volume of a given tissue is similar and reproducible from one subject to another. We hence propose to accelerate modeling of the physical effects occurring in emission tomography making it adequate for daily exam by using the approach of scatter pre-calculated database. The developed efficient patient-dependent attenuation and scatter correction were implemented on a GPU architecture on a state-of-art single-processor workstation and yielded to a speed-up factor in time computing of four orders of magnitude. Results presented in this proof of concept study are in good agreement with full MCS.

A Lagrangian Approach for Calculating Microsphere Deposition in a One-Dimensional Lung-Airway Model.

Using the open-source software openfoam as the solver, a novel approach to calculate microsphere transport and deposition in a 1D human lung-equivalent trumpet model (TM) is presented. Specifically, for particle deposition in a nonlinear trumpetlike configuration a new radial force has been developed which, along with the regular drag force, generates particle trajectories toward the wall. The new semi-empirical force is a function of any given inlet volumetric flow rate, micron-particle diameter, and lung volume. Particle-deposition fractions (DFs) in the size range from 2 μm to 10 μm are in agreement with experimental datasets for different laminar and turbulent inhalation flow rates as well as total volumes. Typical run times on a single processor workstation to obtain actual total deposition results at comparable accuracy are 200 times less than that for an idealized whole-lung geometry (i.e., a 3D-1D model with airways up to 23rd generation in single-path only).

Bloch-Redfield-Wangsness theory engine implementation using symbolic processing software

We describe a general method for the automated symbolic processing of Bloch-Redfield-Wangsness relaxation theory equations for liquid-phase spin dynamics in the algebraically challenging case of rotationally modulated interactions. The processing typically takes no more than a few seconds (on a contemporary single-processor workstation) and yields relaxation rate expressions that are completely general with respect to the spectral density functions, relative orientations, and magnitudes of the interaction tensors, with all cross-correlations accounted for. The algorithm easily deals with fully rhombic interaction tensors, and is able, with little if any modification, to treat a large variety of the relaxation mechanisms encountered in NMR, EPR, and spin dynamics in general.

Mirage: A new iterative map-making code for CMB experiments

A major goal of CMB experiments is to obtain highly sensitive CMB maps in order to extract Spherical Harmonic Power Spectrum (SHPS) and cosmological parameters with unprecedented accuracy. We present a new map-making code (Mirage), based on a two-phase iterative algorithm, involving low frequency drift treatment, Butterworth high-pass filtering and conjugate gradient method. This work was strongly motivated by Archeops CMB experiment data analysis. We believe that Archeops was a good test bench for the future Planck Surveyor data analysis, and Mirage was designed in order to be used for Planck data processing with minimal work. A strong feature of Mirage is that it handles experimental problems in data, such as holes in data stream, bright sources, and galaxy side effects, without jeopardising speed. The other advantage is its processing speed, allowing to run Monte Carlo simulations of Archeops data processing on a single processor workstation overnight. Algorithms are explained. Systematic effects on SHPS are investigated on various simulated data, including typical Archeops observational systematics. This code is available at adress http://www-dapnia.cea.fr/Telechargement/ Preprint with full resolution figures is available at http://www-dapnia.cea.fr/Doc/Publications/Archives/dapnia-03-378.pdf

A smooth particle-mesh Ewald algorithm for Stokes suspension simulations: The sedimentation of fibers

Large-scale simulations of non-Brownian rigid fibers sedimenting under gravity at zero Reynolds number have been performed using a fast algorithm. The mathematical formulation follows the previous simulations by Butler and Shaqfeh [“Dynamic simulations of the inhomogeneous sedimentation of rigid fibres,” J. Fluid Mech. 468, 205 (2002)]. The motion of the fibers is described using slender-body theory, and the line distribution of point forces along their lengths is approximated by a Legendre polynomial in which only the total force, torque, and particle stresslet are retained. Periodic boundary conditions are used to simulate an infinite suspension, and both far-field hydrodynamic interactions and short-range lubrication forces are considered in all simulations. The calculation of the hydrodynamic interactions, which is typically the bottleneck for large systems with periodic boundary conditions, is accelerated using a smooth particle-mesh Ewald (SPME) algorithm previously used in molecular dynamics simulations. In SPME the slowly decaying Green’s function is split into two fast-converging sums: the first involves the distribution of point forces and accounts for the singular short-range part of the interactions, while the second is expressed in terms of the Fourier transform of the force distribution and accounts for the smooth and long-range part. Because of its smoothness, the second sum can be computed efficiently on an underlying grid using the fast Fourier transform algorithm, resulting in a significant speed-up of the calculations. Systems of up to 512 fibers were simulated on a single-processor workstation, providing a different insight into the formation, structure, and dynamics of the inhomogeneities that occur in sedimenting fiber suspensions.

Protein refolding Versus aggregation: computer simulations on an intermediate-resolution protein model

Computer simulations are performed on a system of eight model peptide chains to study how the competition between protein refolding and aggregation affects the optimal conditions for refolding of four-helix bundles. The discontinuous molecular dynamics algorithm is utilized along with an intermediate-resolution protein model that we developed for this work. Physically, the model is much more detailed than any model used to date for simulations of protein aggregation. Each model residue consists of a detailed, three-bead backbone and a simplified, single-bead side-chain. Excluded volume, hydrogen bond, and hydrophobic interactions are modeled with discontinuous (i.e. hard-sphere and square-well) potentials. Simulations efficiently sample conformational space, and complete folding trajectories from random initial configurations to two four-helix bundles are possible within two days on a single processor workstation. Folding of the bundles follows two main pathways, one through a trimeric intermediate and the other through an intermediate with two dimers. The proportion of trajectories that follow each route is significantly different for the eight-peptide system in this work than in a previously studied four-peptide system, which yields one four-helix bundle, suggesting, as our previous simulations have, that protein folding properties are strongly influenced by the presence of other proteins. Folding of the bundles is optimal within a fixed temperature range, with the high-temperature boundary a function of the complexity of the protein (or oligomer) to be folded and the low-temperature boundary a function of the complexity of the protein’s environment. Above the optimal temperature range for folding, the model chains tend to unfold; below the optimal range, the model chains tend to aggregate. As has been seen previously, aggregates have substantial levels of native secondary structure, suggesting that aggregates are composed largely of partially folded intermediates, not denatured chains.

Alpha-helix formation: discontinuous molecular dynamics on an intermediate-resolution protein model.

An intermediate-resolution model of small, homogeneous peptides is introduced, and discontinuous molecular dynamics simulation is applied to study secondary structure formation. Physically, each model residue consists of a detailed three-bead backbone and a simplified single-bead side-chain. Excluded volume and hydrogen bond interactions are constructed with discontinuous (i.e., hard-sphere and square-well) potentials. Simulation results show that the backbone motion of the model is limited to realistic regions of Phi-Psi conformational space. Model polyalanine chains undergo a locally cooperative transition to form alpha-helices that are stabilized by backbone hydrogen bonding, while model polyglycine chains tend to adopt nonhelical structures. When side-chain size is increased beyond a critical diameter, steric interactions prevent formation of long alpha-helices. These trends in helicity as a function of residue type have been well documented by experimental, theoretical, and simulation studies and demonstrate the ability of the intermediate-resolution model developed in this work to accurately mimic realistic peptide behavior. The efficient algorithm used permits observation of the complete helix-coil transition within 15 min on a single-processor workstation, suggesting that simulations of very long times are possible with this model.

Parallel processing for finite-difference modelling of ice sheets

Improved modelling of ice sheets, by use of high resolution and with representation of more physical processes, is constrained by long run-times even on the latest single-processor workstation. Parallel processing therefore has a role to play. This paper describes techniques for the parallel processing of ice sheet models and presents design approaches for both the Cray T3 series and other parallel architectures. An implementation of a fully coupled, thermodynamic, 3D ice sheet model is described for the Cray T3D and is shown to be scaleable and efficient.

Performance Optimization Using Extended Critical Path Analysis in Multithreaded Programs on Multiprocessors

Efficient performance tuning of parallel programs is often hard. Optimization is often done when the program is written as a last effort to increase the performance. With sequential programs each (executed) code segment will affect the completion time. In the case of a parallel program executed on a multiprocessor this is not always true, due to dependencies between the different threads. Thus, certain code segments of the execution may not affect the completion time of the program. Optimization of such code segments will not increase the performance. In this paper we present an approach to optimize performance by finding the extended critical path of the multithreaded program. The extended critical path analysis is a generalization of the critical path analysis in the sense that it also deals with more threads than processors. We have implemented the extended critical path analysis in a performance optimization tool. The tool allows the user to determine the extended critical path of a multithreaded application written for the Solaris operating system for any number of processors based on execution on a single processor workstation.

Optimization of multistage turbines using a through-flow code

Fast and accurate flow calculation and performance prediction of multistage axial flow turbines at design and off-design conditions were performed using a compressible steady state inviscid through-flow code with high fidelity loss and mixing models. The code is based on a stream function model and a finite element solution procedure. A new design system has been developed which optimizes hub and shroud geometry and inlet and exit flow field parameters for each blade row of a multistage axial flow turbine. Optimization was performed using a hybrid constrained optimization code that switches among the modules automatically in order to avoid local minima and to accelerate design convergence rate. By automatically varying a relatively small number of geometric variables per turbine stage it is possible to find an optimal radial distribution of flow parameters at the inlet and outlet of every blade row. Thus, an optimized meridional flow path can be found that is defined by the optimized shape of the hub and shroud while keeping blade shapes intact. The multistage design optimization system has been demonstrated using an actual two-stage axial gas turbine as an example. The comparison of computed performance of an already very high efficiency initial design and its optimized design demonstrates more than 1 per cent improvement in the turbine efficiency at design and significant off-design conditions. The entire design optimization process is feasible on a typical single-processor computer workstation or a personal computer.

New, faster, image-based scatter correction for 3D PET

The authors report on a new numerical implementation of the single-scatter simulation scatter correction algorithm for 3D PET. Its primary advantage over the original implementation is that it is a much faster calculation, currently requiring less than 30 sec execution time per bed position for an adult thorax, thus making clinical whole-body scatter correction more practical. The new code runs on a single processor workstation CPU instead of a vector processor array, making it highly portable. It is modular and independent of any particular reconstruction code. The computed scatter contribution is now intrinsically scaled relative to the emission image and no longer requires normalization to the scatter tails in the sinogram when all activity is contained within the field of view, making it more robust against noise. The new algorithm has been verified against the original code on both phantom and human thorax studies. Initial results indicate that scatter correction may be accurately performed following, instead of prior to, either 3D reprojection or Fourier rebinning. Some evidence is presented that the single-scatter operator, when applied to an uncorrected emission image provides reasonable compensation for multiple scatter.

Ab initio quantum chemistry on PC-based parallel supercomputers

The advent of mass-market personal computers (PC) and the associated price reduction in virtually all computer components has brought the cost of parallel, multi-processor computers down to highly affordable levels. Four-, eight-, and even 12-processor machines, constructed from basic, readily available PC components, can be obtained today for the same price as a good-quality single-processor workstation of a few years ago. Together with now well-established parallel tools (such as the message-passing interface (MPI) or parallel virtual machine (PVM) software), state-of-the-art, fully functioning, parallel machines using the Linux operating system and the latest PC microprocessors can deliver unprecedented price/performance ratios. This article reports on the capabilities and performance of a new, fully parallel ab initio program running on commercially available four- and eight-processor PC-based supercomputers.

Protein docking using spherical polar Fourier correlations

We present a new computational method of docking pairs of proteins by using spherical polar Fourier correlations to accelerate the search for candidate low-energy conformations. Interaction energies are estimated using a hydrophobic excluded volume model derived from the notion of "overlapping surface skins," augmented by a rigorous but "soft" model of electrostatic complementarity. This approach has several advantages over former three-dimensional grid-based fast Fourier transform (FFT) docking correlation methods even though there is no analogue to the FFT in a spherical polar representation. For example, a complete search over all six rigid-body degrees of freedom can be performed by rotating and translating only the initial expansion coefficients, many unfeasible orientations may be eliminated rapidly using only low-resolution terms, and the correlations are easily localized around known binding epitopes when this knowledge is available. Typical execution times on a single processor workstation range from 2 hours for a global search (5 x 10(8) trial orientations) to a few minutes for a local search (over 6 x 10(7) orientations). The method is illustrated with several domain dimer and enzyme-inhibitor complexes and 20 large antibody-antigen complexes, using both the bound and (when available) unbound subunits. The correct conformation of the complex is frequently identified when docking bound subunits, and a good docking orientation is ranked within the top 20 in 11 out of 18 cases when starting from unbound subunits. Proteins 2000;39:178-194.

All-electron ab-initio molecular dynamics

We present an all-electron implementation of the Gaussian and augmented plane wave density functional method (GAPW method), which allows ab-initio density functional calculations for periodic and non-periodic systems. The GAPW method uses a Gaussian basis set to expand the Kohn–Sham orbitals, whereas an augmented plane wave basis set is introduced as an auxiliary basis set to expand the electronic charge density. The results of the all-electron calculations for a representative set of small molecules are reported to demonstrate the accuracy and reliability of the GAPW method. Furthermore, its performance is shown for some larger systems, including calculations on unbranched alkane chains up to n-C100H202 with 1804 basis functions and a fully hydrated crystalline RNA duplex (sodium guanylyl-3′-5′-cytidine nonahydrate) with 368 atoms and 3168 basis functions. Finally, as a first test an all-electron ab-initio molecular dynamics (MD) run was performed for 32 water molecules in a simple cubic box under ambient conditions. A standard single processor workstation (IBM 397) was used for all the presented calculations.

Protein Fold Determination from Sparse Distance Restraints: The Restrained Generic Protein Direct Monte Carlo Method

The authors present the generate-and-select hierarchy for tertiary protein structure prediction. The foundation of this hierarchy is the Restrained Generic Protein (RGP) Direct Monte Carlo method. The RGP method is a highly efficient off-lattice residue buildup procedure that can quickly generate the complete set of topologies that satisfy a very small number of interresidue distance restraints. For three restraints uniformly distributed in a 72-residue protein, the authors demonstrate that the size of this set is {approximately}10{sup 4}. The RGP method can generate this set of structures in less than 1 h using a Silicon Graphics R10000 single processor workstation. Following structure generation, a simple criterion that measures the burial of hydrophobic and hydrophilic residues can reliably select a reduced set of {approximately}10{sup 2} structures that contains the native topology. A minimization of the structures in the reduced set typically ranks the native topology in the five lowest energy folds. Thus, using this hierarchical approach, the authors suggest that de novo prediction of moderate resolution globular protein structure can be achieved in just a few hours on a single processor workstation.

A Defect-Tolerant Computer Architecture: Opportunities for Nanotechnology

Teramac is a massively parallel experimental computer built at Hewlett-Packard Laboratories to investigate a wide range of different computational architectures. This machine contains about 220,000 hardware defects, any one of which could prove fatal to a conventional computer, and yet it operated 100 times faster than a high-end single-processor workstation for some of its configurations. The defect-tolerant architecture of Teramac, which incorporates a high communication bandwith that enables it to easily route around defects, has significant implications for any future nanometer-scale computational paradigm. It may be feasible to chemically synthesize individual electronic components with less than a 100 percent yield, assemble them into systems with appreciable uncertainty in their connectivity, and still create a powerful and reliable data communications network. Future nanoscale computers may consist of extremely large-configuration memories that are programmed for specific tasks by a tutor that locates and tags the defects in the system.

Fusion of images on affine sampling grids

We present a method for combining multiple images of a common object assuming two-dimensional (2D) affine transformations between the image sampling grids. Our method is based upon the projection-onto-convex-sets approach of Yeh Stark [J. Opt. Soc. Am. A7, 491 (1990)]. Each image frame constitutes a single projection in our approach. We derive a frame projection algorithm under the 2D affine transform assumption that uses one-dimensional fast Fourier transform operations. We demonstrate that all the parameters required for successful image fusion can be estimated with sufficient accuracy from the image data proper. Four 64×64-pixel images taken by the Galileo Orbiter spacecraft of the asteroid Gaspra were fused to produce a result with twice the sampling rate and significantly improved spatial resolution. The total processing time was 55 s on a single-processor workstation.

O(N) Scaling Simulations of Silicon Bulk and Surface Properties Based on a Non-Orthogonal Tight-Binding Hamiltonian

AbstractWe have implemented a molecular-dynamics algorithm for silicon using a non-orthogonal tight-binding Hamiltonian with the functional form of Menon and Subbaswamy. Parameters for this Hamiltonian were determined by fitting to a database of first-principles total energy calculations of bulk phases and point defect formation energies. These geometries were chosen to reproduce the configurations seen in defective crystalline and amorphous silicon. We have also implemented the non-orthogonal density-matrix method, paying particular attention to data motion locality to facilitate efficient parallelization of the algorithm. The necessary sparse matrix operations (trace, transpose, matrix multiplication) have also been implemented on a single processor workstation with an algorithm which takes O(N) time. Tests of the method's accuracy involved calculations of surface energies and structural reconstructions and activation energies for bulk diffusion through concerted exchange. We present results of a simulation of the melting and rapid quenching of a silicon sample using molecular-dynamics, and examine the resulting structures.