Overlapping domain decomposition based exponential time differencing methods for semilinear parabolic equations
Not provided.
SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Not provided.
Support Vector Machines (SVM) is a popular machine learning technique, which has been applied to a wide range of domains such as science, finance, and social networks for supervised learning. MaTEx undertakes the challenge of designing a scalable parallel SVM training algorithm for large scale systems, which includes commodity multi-core machines, tightly connected supercomputers and cloud computing systems. Several techniques are proposed for improved speed and memory space usage including adaptive and aggressive elimination of samples for faster convergence , and sparse format representation of data samples. Several heuristics for earliest possible to lazy elimination of non-contributing samples are considered in MaTEx. In many cases, where an early sample elimination might result in a false positive, low overhead mechanisms for reconstruction of key data structures are proposed. The proposed algorithm and heuristics are implemented and evaluated on various publicly available datasets
Support Vector Machines (SVM) is a popular machine learning technique, which has been applied to a wide range of domains such as science, finance, and social networks for supervised learning. MaTEx undertakes the challenge of designing a scalable parallel SVM training algorithm for large scale systems, which includes commodity multi-core machines, tightly connected supercomputers and cloud computing systems. Several techniques are proposed for improved speed and memory space usage including adaptive and aggressive elimination of samples for faster convergence , and sparse format representation of data samples. Several heuristics for earliest possible to lazy elimination of non-contributing samples are considered in MaTEx. In many cases, where an early sample elimination might result in a false positive, low overhead mechanisms for reconstruction of key data structures are proposed. The proposed algorithm and heuristics are implemented and evaluated on various publicly available datasets
Bound states of Quantum Chromodynamics (QCD) give insights into the nature of confinement, a key element of the strong interaction. States may be identified from weak signals extracted from the analysis of high statistics data from reactions with many final state particles. One of the best tools for the analysis of these reactions is Partial Wave Analysis (PWA). PWA transforms an ensemble of experimental data from a large acceptance detector from free particle eigenstates to angular momentum eigenstates. The PWA program must be fast enough to deal with the large amounts of data available currently, as processing time scales with the number of events. The scope of this research is to study the applicability and scalability of Intel's Xeon Phi using the Many Integrated Core (MIC) architecture when applied to the existing PWA code at Jefferson Laboratory. An algorithm was developed for the Xeon Phi and scaled across 240 available threads, giving parallel functionality to the PWA which was originally written serially. This scaling can make the fitting process fifteen times faster.
Concentrated solar power (CSP) plants have the potential to reduce the consumption of nonrenewable resources and greenhouse gas emissions in electricity production. In CSP systems, a field of heliostats focuses solar radiation on a central receiver, and energy is then transferred to a thermal power plant at high temperature. However, maximum receiver surface fluxes are low (30–100 W cm -2 ) with high thermal losses, which has contributed to the limited market penetration of CSP systems. Recently, small (~4 cm 2 ), laminated micro pin-fin devices have shown potential to achieve concentrated surface fluxes over 100 W cm -2 using supercritical CO 2 as the working fluid. The present study explores the feasibility of using these microscale unit cells as building blocks for a megawatt-scale (250 MW thermal) open solar receiver through a numbering-up approach, where multiple microscale unit cell devices are connected in parallel. A multiscale model of the full-scale central receiver is developed. The model consists of interconnected unit cell and module level (i.e., multiple unit cells in parallel) submodels which predict local performance of the central receiver. Each full-scale receiver consists of 3000 micro pin-fin unit cells divided into 250 modules. The performance of three different full-scale receivers is simulated under representative operating conditions. The results show that the microscale unit cells have the potential to be numbered up to megawatt applications while providing high heat flux and thermal efficiency. At the design incident flux and surface emissivity, a global receiver efficiency of approximately 90% when heating sCO 2 from 550 °C to 650 °C at an average incident flux of 110 W cm -2 can be achieved.
We introduce a new benchmark for measuring the performance of parallel input/ouput. This benchmark has flexible initialization. size. and scaling properties that allows it to satisfy seven criteria for practical parallel I/O benchmarks. We obtained performance results while running on the a SGI Origin2OOO computer with various numbers of processors: with 4 processors. the performance was 68.9 Mflop/s with 0.52 of the time spent on I/O, with 8 processors the performance was 139.3 Mflop/s with 0.50 of the time spent on I/O, with 16 processors the performance was 173.6 Mflop/s with 0.43 of the time spent on I/O. and with 32 processors the performance was 259.1 Mflop/s with 0.47 of the time spent on I/O.
An innovative hybrid, analog-digital charge-domain technology, for the massively parallel VLSI implementation of certain large scale matrix-vector operations, has recently been introduced. It employs arrays of Charge Coupled/Charge Injection Device cells holding an analog matrix of charge, which process digital vectors in parallel by means of binary, non-destructive charge transfer operations. The impact of this technology on massively parallel processing is discussed. Fundamentally new classes of algorithms, specifically designed for this emerging technology, as applied to signal processing, are derived.
An innovative hybrid, analog-digital charge-domain technology, for the massively parallel VLSI implementation of certain large scale matrix-vector operations, has recently been introduced. It employs arrays of Charge Coupled/Charge Injection Device cells holding an analog matrix of charge, which process digital vectors in parallel by means of binary, non-destructive charge transfer operations. The impact of this technology on massively parallel processing is discussed.
Abstract Divertor heat flux and its correlation with pedestal parameters within various small edge localized mode (ELM) regimes, including high beta poloidal, type-II and ELMs with negative triangularity H-modes were investigated in DIII-D. The parallel energy fluences of type-II and high beta poloidal small ELM regimes fall below the linear scaling with pedestal electron pressure for type-I ELMs put forward in Eich et al 2017 ( Nucl. Mater. Energy 12 84–90). The negative triangularity of H-mode ELMs follow the Eich scaling for type-I ELMs. The parallel heat flux and total heat loads to the divertor were determined using high-time resolution infrared thermography, while pedestal parameters were obtained through self-consistent kinetic equilibrium reconstructions. Linear regressions for the type-II and high beta poloidal regimes demonstrate that an equivalent 7.5 MA small ELM scenario in ITER would fall below the ~5 MJ m − 2 leading edge melting limit for tungsten (Gunn et al 2017 Nucl. Fusion 57 046025). Utilizing fast thermography, the scrape-off layer power fall-off length for both inter-ELM and intra-ELM was determined and compared to the Eich scaling with poloidal magnetic field in Eich et al (ASDEX Upgrade Team and JET EFDA Contributors 2013 Nucl. Fusion 53 093031). Except for the high beta poloidal scenario, all the small ELM regimes during both inter- and intra-ELM periods had power fall-off lengths ( λ q ) larger then would be expected from the B pol , MP − 1 scaling associated with type-I ELMs, signifying their potential in managing heat loads and offering a solution for core–edge integration.
Matrix-vector multiplication is the key operation in any Krylov-subspace iteration method. We are interested in Krylov methods applied to problems associated with the graph Laplacian arising from large scale-free graphs. Furthermore, computations with graphs of this type on parallel distributed-memory computers are challenging. This is due to the fact that scale-free graphs have a degree distribution that follows a power law, and currently available graph partitioners are not efficient for such an irregular degree distribution. The lack of a good partitioning leads to excessive interprocessor communication requirements during every matrix-vector product. Here, we present an approach to alleviate this problem based on embedding the original irregular graph into a more regular one by disaggregating (splitting up) vertices in the original graph. The matrix-vector operations for the original graph are performed via a factored triple matrix-vector product involving the embedding graph. And even though the latter graph is larger, we are able to decrease the communication requirements considerably and improve the performance of the matrix-vector product.
This paper presents an open-source, parallel AI environment (named OpenGraphGym) to facilitate the application of reinforcement learning (RL) algorithms to address combinatorial graph optimization problems. This environment incorporates a basic deep reinforcement learning method, and several graph embeddings to capture graph features, it also allows users to rapidly plug in and test new RL algorithms and graph embeddings for graph optimization problems. This new open-source RL framework is targeted at achieving both high performance and high quality of the computed graph solutions. This RL framework forms the foundation of several ongoing research directions, including 1) benchmark works on different RL algorithms and embedding methods for classic graph problems; 2) advanced parallel strategies for extreme-scale graph computations, as well as 3) performance evaluation on real-world graph solutions.
Single-Source Shortest Paths (SSSP) is a well-studied graph problem. Examples of SSSP algorithms include the original Dijkstra’s algorithm and the parallel Δ-stepping and KLA-SSSP algorithms. In this paper, we use a novel Abstract Graph Machine (AGM) model to show that all these algorithms share a common logic and differ from one another by the order in which they perform work. We use the AGM model to thoroughly analyze the family of algorithms that arises from the common logic. We start with the basic algorithm without any ordering (Chaotic), and then we derive the existing and new algorithms by methodically exploring semantic and spatial ordering of work. Our experimental results show that new derived algorithms show better performance than the existing distributed memory parallel algorithms, especially at higher scales.
Deterministic sample average approximations of stochastic programming problems with recourse are suitable for a scenario-based parallelization. In this paper the parallelization is obtained by using an interior-point method and a Schur complement mechanism for the interior-point linear systems. However, the direct linear solves involving the dense Schur complement matrix are expensive, and adversely affect the scalability of this approach. We address this issue by proposing a stochastic preconditioner for the Schur complement matrix and by using Krylov iterative methods for the solution of the dense linear systems. The stochastic preconditioner is built based on a subset of existing scenarios and can be assembled and factorized on a separate process before the computation of the Schur complement matrix finishes on the remaining processes. The expensive factorization of the Schur complement is removed from the parallel execution flow and the scaling of the optimization solver is considerably improved with this approach. The spectral analysis indicates an exponentially fast convergence in probability to 1 of the eigenvalues of the preconditioned matrix with the number of scenarios incorporated in the preconditioner. Numerical experiments performed on the relaxation of a unit commitment problem show good performance, in terms of both the accuracy of the solution and the execution time.
In this paper, we propose a novel domain decomposition method for large-scale simulations in continuum mechanics by merging the concepts of Centroidal Voronoi Tessellation (CVT) and Voronoi Particle dynamics (VP). The CVT is introduced to achieve a high-level compactness of the partitioning subdomains by the Lloyd algorithm which monotonically decreases the CVT energy. The number of computational elements between neighboring partitioning subdomains, which scales the communication effort for parallel simulations, is optimized implicitly as the generated partitioning subdomains are convex and simply connected with small aspect-ratios. Moreover, Voronoi Particle dynamics employing physical analogy with a tailored equation of state is developed, which relaxes the particle system towards the target partition with good load balance. Since the equilibrium is computed by an iterative approach, the partitioning subdomains exhibit locality and the incremental property. Numerical experiments reveal that the proposed Centroidal Voronoi Particle (CVP) based algorithm produces high-quality partitioning with high efficiency, independently of computational-element types. Thus it can be used for a wide range of applications in computational science and engineering.
Machine-learning potentials (MLPs) trained on data from quantum-mechanics based first-principles methods can approach the accuracy of the reference method at a fraction of the computational cost. To facilitate efficient MLP-based molecular dynamics and Monte Carlo simulations, an integration of the MLPs with sampling software is needed. Here, we develop two interfaces that link the atomic energy network (ænet) MLP package with the popular sampling packages TINKER and LAMMPS. The three packages, ænet, TINKER, and LAMMPS, are free and open-source software that enable, in combination, accurate simulations of large and complex systems with low computational cost that scales linearly with the number of atoms. Scaling tests show that the parallel efficiency of the ænet–TINKER interface is nearly optimal but is limited to shared-memory systems. The ænet–LAMMPS interface achieves excellent parallel efficiency on highly parallel distributed memory systems and benefits from the highly optimized neighbor list implemented in LAMMPS. We demonstrate the utility of the two MLP interfaces for two relevant example applications: the investigation of diffusion phenomena in liquid water and the equilibration of nanostructured amorphous battery materials.
The magnetic field line random walk (FLRW) is important for the transport of energetic particles in many astrophysical situations. While all authors agree on the quasilinear diffusion of field lines for fluctuations that mainly vary parallel to a large-scale field, for the opposite case of fluctuations that mainly vary in the perpendicular directions, there has been an apparent conflict between concepts of Bohm diffusion and percolation/trapping effects. Here computer simulation and non-perturbative analytic techniques are used to re-examine the FLRW in magnetic turbulence with slab and two-dimensional (2D) components, in which 2D flux surfaces are disturbed by the slab fluctuations. Previous non-perturbative theories for D{sub perpendicular}, based on Corrsin's hypothesis, have identified a slab contribution with quasilinear behavior and a 2D contribution due to Bohm diffusion with diffusive decorrelation (DD), combined in a quadratic formula. Here we present analytic theories for other routes to Bohm diffusion, with random ballistic decorrelation (RBD) either due to the 2D component itself (for a weak slab contribution) or the total fluctuation field (for a strong slab contribution), combined in a direct sum with the slab contribution. Computer simulations confirm the applicability of RBD routes for weak or strong slab contributions, while the DD route applies for a moderate slab contribution. For a very low slab contribution, interesting trapping effects are found, including a depressed diffusion coefficient and subdiffusive behavior. Thus quasilinear, Bohm, and trapping behaviors are all found in the same system, together with an overall viewpoint to explain these behaviors.
In order to study inflationary cosmology and the Milky Way Galaxy's composition and magnetic field structure, Stokes I, Q, and U maps of the Galactic plane covering the Galactic longitude range 260 Degree-Sign < l < 340 Degree-Sign in three atmospheric transmission windows centered on 100, 150, and 220 GHz are presented. The maps sample an optical depth 1 {approx}< A{sub V} {approx}< 30, and are consistent with previous characterizations of the Galactic millimeter-wave frequency spectrum and the large-scale magnetic field structure permeating the interstellar medium. The polarization angles in all three bands are generally perpendicular to those measured by starlight polarimetry as expected and show changes in the structure of the Galactic magnetic field on the scale of 60 Degree-Sign . The frequency spectrum of degree-scale Galactic emission is plotted between 23 and 220 GHz (including WMAP data) and is fit to a two-component (synchrotron and dust) model showing that the higher frequency BICEP data are necessary to tightly constrain the amplitude and spectral index of Galactic dust. Polarized emission is detected over the entire region within two degrees of the Galactic plane, indicating the large-scale magnetic field is oriented parallel to the plane of the Galaxy. A trend of decreasing polarization fraction with increasing total intensity is observed, ruling out the simplest model of a constant Galactic magnetic field orientation along the line of sight in the Galactic plane. A generally increasing trend of polarization fraction with electromagnetic frequency is found, varying from 0.5%-1.5% at frequencies below 50 GHz to 2.5%-3.5% above 90 GHz. The effort to extend the capabilities of BICEP by installing 220 GHz band hardware is described along with analysis of the new band.