Search NASA⌕ Search

SEARCH · Search NASA

Results for “distributed parallelization”

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.

At least 1,423 records · Page 79

Scalable Deep-Learning-Accelerated Topology Optimization for Additively Manufactured Materials

Topology optimization (TO) is a popular and powerful computational approach for designing novel structures, materials, and devices. Two computational challenges have limited the applicability of TO to a variety of industrial applications. First, a TO problem often involves a large number of design variables to guarantee sufficient expressive power. Second, many TO problems require a large number of expensive physical model simulations, and those simulations cannot be parallelized. To address these issues, we propose a general scalable deep-learning (DL) based TO framework, referred to as SDL-TO, which utilizes parallel schemes in high performance computing (HPC) to accelerate the TO process for designing additively manufactured (AM) materials. Unlike the existing studies of DL for TO, our framework accelerates TO by learning the iterative history data and simultaneously training on the mapping between the given design and its gradient. The surrogate gradient is learned by utilizing parallel computing on multiple CPUs incorporated with a distributed DL training on multiple GPUs. The learned TO gradient enables a fast online update scheme instead of an expensive update based on the physical simulator or solver. Using a local sampling strategy, we achieve to reduce the intrinsic high dimensionality of the design space and improve the training accuracy and the scalability of the SDL-TO framework. The method is demonstrated by benchmark examples and AM materials design for heat conduction. The proposed SDL-TO framework shows competitive performance compared to the baseline methods but significantly reduces the computational cost by a speed up of around 8.6x over the standard TO implementation.

Bi, Sirui↗

VVER-440 and VVER-1000 reactor dosimetry benchmark - BUGLE-96 versus ALPAN VII.0

Document available in abstract form only, full text of document follows: Analytical results of the vodo-vodyanoi energetichesky reactor-(VVER-) 440 and VVER-1000 reactor dosimetry benchmarks developed from engineering mockups at the Nuclear Research Inst. Rez LR-0 reactor are discussed. These benchmarks provide accurate determination of radiation field parameters in the vicinity and over the thickness of the reactor pressure vessel. Measurements are compared to calculated results with two sets of tools: TORT discrete ordinates code and BUGLE-96 cross-section library versus the newly Westinghouse-developed RAPTOR-M3G and ALPAN VII.0. The parallel code RAPTOR-M3G enables detailed neutron distributions in energy and space in reduced computational time. ALPAN VII.0 cross-section library is based on ENDF/B-VII.0 and is designed for reactor dosimetry applications. It uses a unique broad group structure to enhance resolution in thermal-neutron-energy range compared to other analogous libraries. The comparison of fast neutron (E > 0.5 MeV) results shows good agreement (within 10%) between BUGLE-96 and ALPAN VII.O libraries. Furthermore, the results compare well with analogous results of participants of the REDOS program (2005). Finally, the analytical results for fast neutrons agree within 15% with the measurements, for most locations in all three mockups. In general, however, the analytical results underestimate the attenuation through the reactor pressure vessel thickness compared to the measurements. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

High resolution river routing in the Community Climate System Model

The current version of the Community Climate System Model CCSM uses half degree resolution river routing within the land component of CCSM. We present a scaling approach and status on a project to produce a much higher resolution data set for river routing to go along with higher resolution land cover data sets for the Community Land Model CLM in order to take advantage of the increasing computational power now available. The new higher resolution data set is based on the Hydrosheds and Hydro1K datasets from USGS. The flow directions are used to generate basins so that the computational load can be distributed among processors by basins to minimize the parallel communication necessary. The code modifications will make the river component more scalable and efficient. The higher resolution models enable detailed study of climatic effects from human induced land cover/land use changes such as the deployment of biofuel crops for energy production.

99 GENERAL AND MISCELLANEOUS↗

Reconciling Scratch Space Consumption, Exposure, and Volatility to Achieve Timely Staging of Job Input Data

Innovative scientific applications and emerging dense data sources are creating a data deluge for high-end computing systems. Processing such large input data typically involves copying (or staging) onto the supercomputer's specialized high-speed storage, scratch space, for sustained high I/O throughput. The current practice of conservatively staging data as early as possible makes the data vulnerable to storage failures, which may entail re-staging and consequently reduced job throughput. To address this, we present a timely staging framework that uses a combination of job startup time predictions, user-specified intermediate nodes, and decentralized data delivery to coincide input data staging with job start-up. By delaying staging to when it is necessary, the exposure to failures and its effects can be reduced. Evaluation using both PlanetLab and simulations based on three years of Jaguar (No. 1 in Top500) job logs show as much as 85.9% reduction in staging times compared to direct transfers, 75.2% reduction in wait time on scratch, and 2.4% reduction in usage/hour.

99 GENERAL AND MISCELLANEOUS↗

Software Engineering Support of the Third Round of Scientific Grand Challenge Investigations: Earth System Modeling Software Framework Survey

One of the most significant challenges in large-scale climate modeling, as well as in high-performance computing in other scientific fields, is that of effectively integrating many software models from multiple contributors. A software framework facilitates the integration task, both in the development and runtime stages of the simulation. Effective software frameworks reduce the programming burden for the investigators, freeing them to focus more on the science and less on the parallel communication implementation. while maintaining high performance across numerous supercomputer and workstation architectures. This document surveys numerous software frameworks for potential use in Earth science modeling. Several frameworks are evaluated in depth, including Parallel Object-Oriented Methods and Applications (POOMA), Cactus (from (he relativistic physics community), Overture, Goddard Earth Modeling System (GEMS), the National Center for Atmospheric Research Flux Coupler, and UCLA/UCB Distributed Data Broker (DDB). Frameworks evaluated in less detail include ROOT, Parallel Application Workspace (PAWS), and Advanced Large-Scale Integrated Computational Environment (ALICE). A host of other frameworks and related tools are referenced in this context. The frameworks are evaluated individually and also compared with each other.

Talbot, Bryan↗

Parallel FETI‐DP algorithm for efficient simulation of large‐scale EM problems

Summary An efficient parallelization of the dual‐primal finite‐element tearing and interconnecting (FETI‐DP) algorithm is presented for large‐scale electromagnetic simulations. As a nonoverlapping domain decomposition method, the FETI‐DP algorithm formulates a global interface problem, whose iterative solution is accelerated with a solution of a global corner problem. To achieve a good load balance for parallel computation, the original computational domain is decomposed into subdomains with similar sizes and shapes. The subdomains are then distributed to processors based on their close proximity to minimize inter‐processor communication. The parallel generalized minimal residual method, enhanced with the iterative classical Gram‐Schmidt orthogonalization scheme to reduce global communication, is adopted to solve the global interface problem with a fast convergence rate. The global corner‐related coarse problem is solved iteratively with a parallel communication‐avoiding biconjugate gradient stabilized method to minimize global communication, and its convergence is accelerated by a diagonal preconditioner constructed from the coarse system matrix. To alleviate neighboring communication overhead, the non‐blocking communication approach is employed in both generalized minimal residual and communication‐avoiding biconjugate gradient stabilized iterative solutions. Three numerical examples are presented to demonstrate the accuracy, scalability, and capability of the proposed parallel FETI‐DP algorithm for electromagnetic modeling of general objects and antenna arrays. Copyright © 2016 John Wiley & Sons, Ltd.

Zhang, Kedi↗

Mr. Scan: extreme scale density-based clustering using a tree-based network of GPGPU nodes, In: SC '13 Proceedings of the International Conference on High Performance Computing, Networking, Storage and Analysis

Density-based clustering algorithms are a widely-used class of data mining techniques that can find irregularly shaped clusters and cluster data without prior knowledge of the number of clusters it contains. DBSCAN is the most well-known density-based clustering algorithm. We introduce our version of DBSCAN, called Mr. Scan, which uses a hybrid parallel implementation that combines the MRNet tree-based distribution network with GPGPU-equipped nodes. Mr. Scan avoids the problems of existing implementations by effectively partitioning the point space and by optimizing DBSCAN's computation over dense data regions. We tested Mr. Scan on both a geolocated Twitter dataset and image data obtained from the Sloan Digital Sky Survey. At its largest scale, Mr. Scan clustered 6.5 billion points from the Twitter dataset on 8,192 GPU nodes on Cray Titan in 17.3 minutes. All other parallel DBSCAN implementations have only demonstrated the ability to cluster up to 100 million points.

Computer Science↗

High Gradient High Shunt Impedance Accelerating Structure

develop an inexpensive accelerating structure with a high gradient and high shunt impedance. Euclid Techlabs proposes to optimize this technology (based on a successful accelerating structure with distributed coupling developed at SLAC) for industrial applications and serial production. Future colliders require high gradient and high efficiency accelerating structures (high shunt impedance). The SLAC accelerator laboratory recently developed a series of accelerating structures with distributed coupling. Such a structure includes one or more parallel waveguides which are loaded by accelerating cavities. This approach allows configurations where no RF power is flowing axially through the accelerating cavity while maintaining a traveling RF wave through the cross-section of the accelerating structure. In this case the accelerating cavities are isolated at the beam pipe. Conventional traveling wave structures typically couple the RF power through the beam pipe. This feature constrains the cavity shunt impedance to relatively small values. For this reason, the distributed coupling design improves efficiency while reducing cost and improving the operational flexibility of particle accelerators. High gradient operation poses the risk of having a high breakdown rates. In traditional standing and traveling wave structures, RF power flows through the accelerating cells. For distributed coupling structures, this risk is reduced because the absence of power flow through the accelerating cavity allows configurations where no power flows through the beam pipe. Finally, the distributed coupling accelerating structure design allows fabricating structures in two halves. This significantly reduces the complexity of machining, brazing and tuning.

43 PARTICLE ACCELERATORS↗

COLD FRONTS AND GAS SLOSHING IN GALAXY CLUSTERS WITH ANISOTROPIC THERMAL CONDUCTION

Cold fronts in cluster cool cores should be erased on short timescales by thermal conduction, unless protected by magnetic fields that are 'draped' parallel to the front surfaces, suppressing conduction perpendicular to the sloshing fronts. We present a series of MHD simulations of cold front formation in the core of a galaxy cluster with anisotropic thermal conduction, exploring a parameter space of conduction strengths parallel and perpendicular to the field lines. Including conduction has a strong effect on the temperature distribution of the core and the appearance of the cold fronts. Though magnetic field lines are draping parallel to the front surfaces, preventing conduction directly across them, the temperature jumps across the fronts are nevertheless reduced. The geometry of the field is such that the cold gas below the front surfaces can be connected to hotter regions outside via field lines along directions perpendicular to the plane of the sloshing motions and along sections of the front that are not perfectly draped. This results in the heating of this gas below the front on a timescale of a Gyr, but the sharpness of the density and temperature jumps may nevertheless be preserved. By modifying the gas density distribution below the front, conduction may indirectly aid in suppressing Kelvin-Helmholtz instabilities. If conduction along the field lines is unsuppressed, we find that the characteristic sharp jumps seen in Chandra observations of cold front clusters do not form. Therefore, the presence of cold fronts in hot clusters is in contradiction with our simulations with full Spitzer conduction. This suggests that the presence of cold fronts in hot clusters could be used to place upper limits on conduction in the bulk of the intracluster medium. Finally, the combination of sloshing and anisotropic thermal conduction can result in a larger flux of heat to the core than either process in isolation. While still not sufficient to prevent a cooling catastrophe in the very central (r {approx} 5 kpc) regions of the cool core (where something else is required, such as active galactic nucleus feedback), it reduces significantly the mass of gas that experiences a cooling catastrophe outside those small radii.

79 ASTRONOMY AND ASTROPHYSICS↗

Monte Carlo calculations of correction factors for plane-parallel ionization chambers in clinical electron dosimetry

Recent standard dosimetry protocols recommend that plane-parallel ionization chambers be used in the measurements of depth-dose distributions or the calibration of low-energy electron beams with beam quality R{sub 50}<4 g/cm{sup 2}. In electron dosimetry protocols with the plane-parallel chambers, the wall correction factor, P{sub wall}, in water is assumed to be unity and the replacement correction factor, P{sub repl}, is taken to be unity for well-guarded plane-parallel chambers, at all measurement depths. This study calculated P{sub wall} and P{sub repl} for NACP-02, Markus, and Roos plane-parallel chambers in clinical electron dosimetry using the EGSnrc Monte Carlo code system. The P{sub wall} values for the plane-parallel chambers increased rapidly as a function of depth in water, especially at lower energy. The value around R{sub 50} for NACP-02 was about 10% greater than unity at 4 MeV. The effect was smaller for higher electron energies. Similarly, P{sub repl} values with depth increased drastically at the region with the steep dose gradient for lower energy. For Markus P{sub repl} departed more than 10% from unity close to R{sub 50} due to the narrow guard ring width. P{sub repl} for NACP-02 and Roos was close to unity in the plateau region of depth-dose curves that includes a reference depth, d{sub ref}. It was also found that the ratio of the dose to water and the dose to the sensitive volume in the air cavity for the plane-parallel chambers, D{sub w}/[D{sub air}]{sub pp}, at d{sub ref} differs significantly from that assumed by electron dosimetry protocols.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

RADIAL EVOLUTION OF THE WAVEVECTOR ANISOTROPY OF SOLAR WIND TURBULENCE BETWEEN 0.3 AND 1 AU

We present observations of the power spectral anisotropy in the wavevector space of solar wind turbulence and study how it evolves in interplanetary space with increasing heliocentric distance. We use magnetic field measurements from the Helios 2 spacecraft within 1 AU. To derive the power spectral density (PSD) in the (k{sub Parallel-To }, k ) space based on single-satellite measurements is a challenging task that had not been accomplished previously. Here, we derive the spectrum PSD{sub 2D}(k{sub Parallel-To }, k{sub Up-Tack }) from the spatial correlation function CF{sub 2D}(r{sub Parallel-To }, r ) by a transformation according to the projection-slice theorem. We find the so-constructed PSDs to be distributed in k space mainly along a ridge that is more inclined toward the k{sub Up-Tack} axis than the k{sub Parallel-To} axis. Furthermore, this ridge of the distribution is found to gradually get closer to the k{sub Up-Tack} axis as the outer scale length of the turbulence becomes larger with increasing radial distance. In the vicinity of the k{sub Parallel-To} axis, a minor spectral component appears that probably corresponds to quasi-parallel Alfvenic fluctuations. Their relative contribution to the total spectral density tends to decrease with radial distance. These findings suggest that solar wind turbulence undergoes an anisotropic cascade transporting most of its magnetic energy toward larger k{sub Up-Tack} and that the anisotropy in the inertial range is radially developing further at scales that are relatively far from the ever increasing outer scale. For the ion-scale fluctuations, we speculate, from the radial evolution of the extended oblique major component, a transition tendency from dominance by oblique Alfven/ion-cyclotron waves (<1 AU) to dominance by kinetic Alfven waves (>1 AU)

79 ASTRONOMY AND ASTROPHYSICS↗

Calculation of Pressure Distribution on Airship Hulls

These calculations were based on the shape of the ZR III, with the following simplifications: cars, fins, and rudders removed; all cross sections replaced by equivalent circular cross sections. Under these assumptions the pressure distribution was calculated for the following cases: symmetrical case, or flow parallel to the axis; unsymmetrical case, or flow at an angle to the axis. In both cases the simple potential flow first forms the basis for the determination of the pressure distribution.

Von Karman, Theodor↗

A Navier-Strokes Chimera Code on the Connection Machine CM-5: Design and Performance

We have implemented a three-dimensional compressible Navier-Stokes code on the Connection Machine CM-5. The code is set up for implicit time-stepping on single or multiple structured grids. For multiple grids and geometrically complex problems, we follow the 'chimera' approach, where flow data on one zone is interpolated onto another in the region of overlap. We will describe our design philosophy and give some timing results for the current code. A parallel machine like the CM-5 is well-suited for finite-difference methods on structured grids. The regular pattern of connections of a structured mesh maps well onto the architecture of the machine. So the first design choice, finite differences on a structured mesh, is natural. We use centered differences in space, with added artificial dissipation terms. When numerically solving the Navier-Stokes equations, there are liable to be some mesh cells near a solid body that are small in at least one direction. This mesh cell geometry can impose a very severe CFL (Courant-Friedrichs-Lewy) condition on the time step for explicit time-stepping methods. Thus, though explicit time-stepping is well-suited to the architecture of the machine, we have adopted implicit time-stepping. We have further taken the approximate factorization approach. This creates the need to solve large banded linear systems and creates the first possible barrier to an efficient algorithm. To overcome this first possible barrier we have considered two options. The first is just to solve the banded linear systems with data spread over the whole machine, using whatever fast method is available. This option is adequate for solving scalar tridiagonal systems, but for scalar pentadiagonal or block tridiagonal systems it is somewhat slower than desired. The second option is to 'transpose' the flow and geometry variables as part of the time-stepping process: Start with x-lines of data in-processor. Form explicit terms in x, then transpose so y-lines of data are in-processor. Form explicit terms in y, then transpose so z-lines are in processor. Form explicit terms in z, then solve linear systems in the z-direction. Transpose to the y-direction, then solve linear systems in the y-direction. Finally transpose to the x direction and solve linear systems in the x-direction. This strategy avoids inter-processor communication when differencing and solving linear systems, but requires a large amount of communication when doing the transposes. The transpose method is more efficient than the non-transpose strategy when dealing with scalar pentadiagonal or block tridiagonal systems. For handling geometrically complex problems the chimera strategy was adopted. For multiple zone cases we compute on each zone sequentially (using the whole parallel machine), then send the chimera interpolation data to a distributed data structure (array) laid out over the whole machine. This information transfer implies an irregular communication pattern, and is the second possible barrier to an efficient algorithm. We have implemented these ideas on the CM-5 using CMF (Connection Machine Fortran), a data parallel language which combines elements of Fortran 90 and certain extensions, and which bears a strong similarity to High Performance Fortran. We make use of the Connection Machine Scientific Software Library (CMSSL) for the linear solver and array transpose operations.

Jespersen, Dennis C.↗

Parallel simulated annealing algorithms for cell placement on hypercube multiprocessors

Two parallel algorithms for standard cell placement using simulated annealing are developed to run on distributed-memory message-passing hypercube multiprocessors. The cells can be mapped in a two-dimensional area of a chip onto processors in an n-dimensional hypercube in two ways, such that both small and large cell exchange and displacement moves can be applied. The computation of the cost function in parallel among all the processors in the hypercube is described, along with a distributed data structure that needs to be stored in the hypercube to support the parallel cost evaluation. A novel tree broadcasting strategy is used extensively for updating cell locations in the parallel environment. A dynamic parallel annealing schedule estimates the errors due to interacting parallel moves and adapts the rate of synchronization automatically. Two novel approaches in controlling error in parallel algorithms are described: heuristic cell coloring and adaptive sequence control.

Banerjee, Prithviraj↗

Electron dynamics in two-dimensional asymmetric anti-parallel reconnection

Kinetic simulations and spacecraft observations have documented strong anisotropy in the electron distribution function during magnetic reconnection. The level and role of electron pressure anisotropy is investigated for asymmetric geometries applicable to reconnection in the day-side magnetopause. A previously derived analytic model for the pressure anisotropy is generalized and is applied to the asymmetric geometry. In agreement with the results from a kinetic simulation, the generalized model predicts the strongest pressure anisotropy and parallel electric fields in the in-flow region characterized by low electron pressure.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Dynamics of a reconnection-driven runaway ion tail in a reversed field pinch plasma

While reconnection-driven ion heating is common in laboratory and astrophysical plasmas, the underlying mechanisms for converting magnetic to kinetic energy remain not fully understood. Reversed field pinch discharges are often characterized by rapid ion heating during impulsive reconnection, generating an ion distribution with an enhanced bulk temperature, mainly perpendicular to magnetic field. In the Madison Symmetric Torus, a subset of discharges with the strongest reconnection events develop a very anisotropic, high energy tail parallel to magnetic field in addition to bulk perpendicular heating, which produces a fusion neutron flux orders of magnitude higher than that expected from a Maxwellian distribution. Here, we demonstrate that two factors in addition to a perpendicular bulk heating mechanism must be considered to explain this distribution. First, ion runaway can occur in the strong parallel-to-B electric field induced by a rapid equilibrium change triggered by reconnection-based relaxation; this effect is particularly strong on perpendicularly heated ions which experience a reduced frictional drag relative to bulk ions. Second, the confinement of ions varies dramatically as a function of velocity. Whereas thermal ions are governed by stochastic diffusion along tearing-altered field lines (and radial diffusion increases with parallel speed), sufficiently energetic ions are well confined, only weakly affected by a stochastic magnetic field. High energy ions traveling mainly in the direction of toroidal plasma current are nearly classically confined, while counter-propagating ions experience an intermediate confinement, greater than that of thermal ions but significantly less than classical expectations. The details of ion confinement tend to reinforce the asymmetric drive of the parallel electric field, resulting in a very asymmetric, anisotropic distribution.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Computationally efficient implementation of combustion chemistry in parallel PDF calculations

In parallel calculations of combustion processes with realistic chemistry, the serial in situ adaptive tabulation (ISAT) algorithm [S.B. Pope, Computationally efficient implementation of combustion chemistry using in situ adaptive tabulation, Combustion Theory and Modelling, 1 (1997) 41-63; L. Lu, S.B. Pope, An improved algorithm for in situ adaptive tabulation, Journal of Computational Physics 228 (2009) 361-386] substantially speeds up the chemistry calculations on each processor. To improve the parallel efficiency of large ensembles of such calculations in parallel computations, in this work, the ISAT algorithm is extended to the multi-processor environment, with the aim of minimizing the wall clock time required for the whole ensemble. Parallel ISAT strategies are developed by combining the existing serial ISAT algorithm with different distribution strategies, namely purely local processing (PLP), uniformly random distribution (URAN), and preferential distribution (PREF). The distribution strategies enable the queued load redistribution of chemistry calculations among processors using message passing. They are implemented in the software x2f{sub m}pi, which is a Fortran 95 library for facilitating many parallel evaluations of a general vector function. The relative performance of the parallel ISAT strategies is investigated in different computational regimes via the PDF calculations of multiple partially stirred reactors burning methane/air mixtures. The results show that the performance of ISAT with a fixed distribution strategy strongly depends on certain computational regimes, based on how much memory is available and how much overlap exists between tabulated information on different processors. No one fixed strategy consistently achieves good performance in all the regimes. Therefore, an adaptive distribution strategy, which blends PLP, URAN and PREF, is devised and implemented. It yields consistently good performance in all regimes. In the adaptive parallel ISAT strategy, the type and extent of redistribution is determined 'on the fly' based on the prediction of future simulation time. Compared to the PLP/ISAT strategy where chemistry calculations are essentially serial, a speed-up factor of up to 30 is achieved. The study also demonstrates that the adaptive strategy has acceptable parallel scalability.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

A Case for Optimistic Coordination in HPC Storage Systems

High-performance computing (HPC) storage systems rely on access coordination to ensure that concurrent updates do not produce incoherent results. HPC storage systems typically employ pessimistic distributed locking to provide this functionality in cases where applications cannot perform their own coordination. This approach, however, introduces significant performance overhead and complicates fault handling. In this work we evaluate the viability of optimistic conditional storage operations as an alternative to distributed locking in HPC storage systems. We investigate design strategies and compare the two approaches in a prototype object storage system using a parallel read/modify/write benchmark. Our prototype illustrates that conditional operations can be easily integrated into distributed object storage systems and can outperform standard coordination primitives for simple update workloads. Our experiments show that conditional updates can achieve over two orders of magnitude higher performance than pessimistic locking for some parallel read/modify/write workloads.

Carns, Philip↗