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 1,405 records · Page 78

Evidence of topological charge polarization at active-passive interfaces in acoustically powered active liquid crystals

Theoretical and computational studies predict topological charge separation at active-passive interfaces in active nematics, but reliable experimental validation has been lacking. We utilize a synthetic acoustically energized liquid crystal to experimentally investigate the dynamics of topological defects across an active-passive interface. The interference pattern induced by the acoustic wave inside the experimental cell creates a spatial distribution of activity, resulting in the formation of active-passive interfaces with the liquid-crystalline director field aligned parallel to those interfaces. The activity gradient drives the reorientation of positively charged topological defects along its direction, causing them to migrate into the passive zone and form a layer with net topological charge. That layer “discharges” upon cessation of activity through annihilation of positive and negative topological charges, resembling the discharge of a capacitor.

Active defects↗

The Reverse Cuthill-McKee Algorithm in Distributed-Memory

Ordering vertices of a graph is key to minimize fill-in and data structure size in sparse direct solvers, maximize locality in iterative solvers, and improve performance in graph algorithms. Except for naturally parallelizable ordering methods such as nested dissection, many important ordering methods have not been efficiently mapped to distributed-memory architectures. In this paper, we present the first-ever distributed-memory implementation of the reverse Cuthill-McKee (RCM) algorithm for reducing the profile of a sparse matrix. Our parallelization uses a two-dimensional sparse matrix decomposition. We achieve high performance by decomposing the problem into a small number of primitives and utilizing optimized implementations of these primitives. Our implementation shows strong scaling up to 1024 cores for smaller matrices and up to 4096 cores for larger matrices.

97 MATHEMATICS AND COMPUTING↗

A Communication-Avoiding 3D LU Factorization Algorithm for Sparse Matrices

We propose a new algorithm to improve the strong scalability of right-looking sparse LU factorization on distributed memory systems. Our 3D sparse LU algorithm uses a three-dimensional MPI process grid, aggressively exploits elimination tree parallelism and trades off increased memory for reduced per-process communication. We also analyze the asymptotic improvements for planar graphs (e.g., from 2D grid or mesh domains) and certain non-planar graphs (specifically for 3D grids and meshes). For planar graphs with n vertices, our algorithm reduces communication volume asymptotically in n by a factor of O{log n} and latency by a factor of O{log n}. For non-planar cases, our algorithm can reduce the per-process communication volume by 3× and latency by O{n^1/3} times. In all cases, the memory needed to achieve these gains is a constant factor. We implemented our algorithm by extending the 2D data structure used in superLU. Our new 3D code achieves speedups up to 27× for planar graphs and up to 3.3× for non-planar graphs over the baseline 2D superLU when run on 24,000 cores of a Cray XC30.

Sao, Piyush↗

Knowledge Discovery for Smart Grid Operation, Control, and Situation Awareness -- A Big Data Visualization Platform

In this paper, a big data visualization platform is designed to discover the hidden useful knowledge for smart grid (SG) operation, control and situation awareness. The spawn of smart sensors at both grid side and customer side can provide large volume of heterogeneous data that collect information in all time spectrums. Extracting useful knowledge from this big-data poll is still challenging. In this paper, the Apache Spark, an open source cluster computing framework, is used to process the big-data to effectively discover the hidden knowledge. A high-speed communication architecture utilizing the Open System Interconnection (OSI) model is designed to transmit the data to a visualization platform. This visualization platform uses Google Earth, a global geographic information system (GIS) to link the geological information with the SG knowledge and visualize the information in user defined fashion. The University of Denver's campus grid is used as a SG test bench and several demonstrations are presented for the proposed platform.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Stoichiometry of microbial carbon use efficiency in soils

The carbon use efficiency (CUE) of microbial communities partitions the flow of C from primary producers to the atmosphere, decomposer food webs, and soil C stores. CUE, usually defined as the ratio of growth to assimilation, is a critical parameter in ecosystem models, but is seldom measured directly in soils because of the methodological difficulty of measuring in situ rates of microbial growth and respiration. Alternatively, CUE can be estimated indirectly from the elemental stoichiometry of organic matter and microbial biomass, and the ratios of C to nutrient-acquiring ecoenzymatic activities. In this paper, we used this approach to estimate and compare microbial CUE in >2000 soils from a broad range of ecosystems. Mean CUE based on C:N stoichiometry was 0.269 ± 0.110 (mean ± SD). A parallel calculation based on C:P stoichiometry yielded a mean CUE estimate of 0.252 ± 0.125. The mean values and frequency distributions were similar to those from aquatic ecosystems, also calculated from stoichiometric models, and to those calculated from direct measurements of bacterial and fungal growth and respiration. CUE was directly related to microbial biomass C with a scaling exponent of 0.304 (95% CI 0.237–0.371) and inversely related to microbial biomass P with a scaling exponent of -0.234 (95% CI -0.289 to -0.179). Relative to CUE, biomass specific turnover time increased with a scaling exponent of 0.509 (95% CI 0.467–0.551). CUE increased weakly with mean annual temperature. CUE declined with increasing soil pH reaching a minimum at pH 7.0, then increased again as soil pH approached 9.0, a pattern consistent with pH trends in the ratio of fungal : bacteria abundance and growth. Structural equation models that related geographic variables to CUE component variables showed the strongest connections for paths linking latitude and pH to β-glucosidase activity and soil C:N:P ratios. Finally, the integration of stoichiometric and metabolic models provides a quantitative description of the functional organization of soil microbial communities that can improve the representation of CUE in microbial process and ecosystem simulation models.

54 ENVIRONMENTAL SCIENCES↗

Efficient Support for Matrix Computations on Heterogeneous Multi-core and Multi-GPU Architectures

We present a new methodology for utilizing all CPU cores and all GPUs on a heterogeneous multicore and multi-GPU system to support matrix computations e ciently. Our approach is able to achieve the objectives of a high degree of parallelism, minimized synchronization, minimized communication, and load balancing. Our main idea is to treat the heterogeneous system as a distributed-memory machine, and to use a heterogeneous 1-D block cyclic distribution to allocate data to the host system and GPUs to minimize communication. We have designed heterogeneous algorithms with two di erent tile sizes (one for CPU cores and the other for GPUs) to cope with processor heterogeneity. We propose an auto-tuning method to determine the best tile sizes to attain both high performance and load balancing. We have also implemented a new runtime system and applied it to the Cholesky and QR factorizations. Our experiments on a compute node with two Intel Westmere hexa-core CPUs and three Nvidia Fermi GPUs demonstrate good weak scalability, strong scalability, load balance, and e ciency of our approach.

97 MATHEMATICS AND COMPUTING↗

Transfer Function Design for Scientific Discovery

As computation scales beyond terascale, the scientific problems under study through computing are increasingly pushing the boundaries of human knowledge about the physical world. It is more pivotal than ever to quickly and reliably extract new knowledge from these complex simulations of ultra scale. In this project, the PI expanded the traditional notion of transfer function, which maps physical quantities to visual cues via table look-ups, to include general temporal as well as multivariate patterns that can be described procedurally through specialty mini programming languages. Their efforts aimed at answering a perpetual question of fundamental importance. That is "what a visualization should show". Instead of waiting for application scientists to initiate the process, the team at University of Tennessee worked closely with scientists at ORNL in a proactive role to envision and design elegant, powerful, and reliable tools that a user can use to specify "what is interesting". Their new techniques include visualization operators that revolve around correlation and graph properties, relative patterns in statistical distribution, temporal regular expressions, concurrent attribute subspaces and traditional compound boolean range queries. The team also paid special attention to ensure that all visualization operators are inherently designed with great parallel scalability to handle tera-scale datasets in both homogeneous and heterogeneous environments. Success has been demonstrated with leading edge computational science areas include climate modeling, combustion and systems genetics.

97 MATHEMATICS AND COMPUTING↗

A WRF simulation of the impact of 3-D radiative transfer on surface hydrology over the Rocky Mountains and Sierra Nevada

We investigate 3-D mountains/snow effects on solar flux distributions and their impact on surface hydrology over the western United States, specifically the Rocky Mountains and Sierra Nevada. The Weather Research and Forecasting (WRF) model, applied at a 30 km grid resolution, is used in conjunction with a 3-D radiative transfer parameterization covering a time period from 1 November 2007 to 31 May 2008, during which abundant snowfall occurred. A comparison of the 3-D WRF simulation with the observed snow water equivalent (SWE) and precipitation from Snowpack Telemetry (SNOTEL) sites shows reasonable agreement in terms of spatial patterns and daily and seasonal variability, although the simulation generally has a positive precipitation bias. We show that 3-D mountain features have a profound impact on the diurnal and monthly variation of surface radiative and heat fluxes, and on the consequent elevation-dependence of snowmelt and precipitation distributions. In particular, during the winter months, large deviations (3-D-PP, in which PP denotes the plane-parallel approach) of the monthly mean surface solar flux are found in the morning and afternoon hours due to shading effects for elevations below 2.5 km. During spring, positive deviations shift to the earlier morning. Over mountaintops higher than 3 km, positive deviations are found throughout the day, with the largest values of 40–60 W m -2 occurring at noon during the snowmelt season of April to May. The monthly SWE deviations averaged over the entire domain show an increase in lower elevations due to reduced snowmelt, which leads to a reduction in cumulative runoff. Over higher elevation areas, positive SWE deviations are found because of increased solar radiation available at the surface. Overall, this study shows that deviations of SWE due to 3-D radiation effects range from an increase of 18% at the lowest elevation range (1.5–2 km) to a decrease of 8% at the highest elevation range (above 3 km). Since lower elevation areas occupy larger fractions of the land surface, the net effect of 3-D radiative transfer is to extend snowmelt and snowmelt-driven runoff into the warm season.Finally, because 60–90% of water resources originate from mountains worldwide, the aforementioned differences in simulated hydrology due solely to 3-D interactions between solar radiation and mountains/snow merit further investigation in order to understand the implications of modeling mountain water resources, and these resources' vulnerability to climate change and air pollution.

3-D radiative transfer↗

A Divergence Statistics Extension to VTK for Performance Analysis

This report follows the series of previous documents ([PT08, BPRT09b, PT09, BPT09, PT10, PB13], where we presented the parallel descriptive, correlative, multi-correlative, principal component analysis, contingency, k -means, order and auto-correlative statistics engines which we developed within the Visualization Tool Kit ( VTK ) as a scalable, parallel and versatile statistics package. We now report on a new engine which we developed for the calculation of divergence statistics, a concept which we hereafter explain and whose main goal is to quantify the discrepancy, in a stasticial manner akin to measuring a distance, between an observed empirical distribution and a theoretical, "ideal" one. The ease of use of the new diverence statistics engine is illustrated by the means of C++ code snippets. Although this new engine does not yet have a parallel implementation, it has already been applied to HPC performance analysis, of which we provide an example.

97 MATHEMATICS AND COMPUTING↗

Sodium-cooled Fast Reactor (SFR) Coolant Flow

Pressure distribution in a 217-pin fuel bundle with wire-wrapped spacers. The Nek5000 simulation calculations were run with 3 million spectral elements of order N=7 (one billion points total) with sustained 80% parallel efficiency. Credits; Science: Paul Fischer, Argonne National Laboratory. Visualization: Hank Childs, Lawrence Berkeley National Laboratory

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Maximal clique enumeration with data-parallel primitives

The enumeration of all maximal cliques in an undirected graph is a fundamental problem arising in several research areas. We consider maximal clique enumeration on shared-memory, multi-core architectures and introduce an approach consisting entirely of data-parallel operations, in an effort to achieve efficient and portable performance across different architectures. We study the performance of the algorithm via experiments varying over benchmark graphs and architectures. Overall, we observe that our algorithm achieves up to a 33-time speedup and 9-time speedup over state-of-the-art distributed and serial algorithms, respectively, for graphs with higher ratios of maximal cliques to total cliques. Further, we attain additional speedups on a GPU architecture, demonstrating the portable performance of our data-parallel design.

97 MATHEMATICS AND COMPUTING↗

Nonlinear simulations of GAEs in NSTX-U

A set of nonlinear simulations has been performed in order to study the nonlinear evolution of unstable global Alfvén eigenmodes in the National Spherical Torus Experiment-Upgrade (NSTX-U). Results of the single toroidal mode number, n, simulations are compared with a full nonlinear simulation (all toroidal harmonics included). In single-n simulations, the conservation of two integrals of motion of a particle in a cyclotron resonance with a monochromatic wave is demonstrated, resulting in a one-dimensional evolution of the particle distribution in (E,μ,pϕ) phase-space. Nonlinear simulations (both single-n and full nonlinear) show a significant redistribution of the resonant fast ions, especially in the pitch parameter. Thus, the changes in the resonant particle's parallel and perpendicular energies can be several times larger than the total particle energy change, with only a small fraction transferring into the excitation of the mode itself. This implies that even a relatively small amplitude mode can significantly modify the beam distribution in the resonant region. For the NSTX-U case considered, the single-n simulation results are close to full nonlinear simulation only for the most unstable mode, in which case the saturation amplitudes and changes in the fast ion distribution are comparable. In contrast, peak amplitudes of subdominant modes in all-n simulations are smaller by a factor of 3–10 compared to single-n runs due to the flattening of the beam ion distribution by the fastest growing mode.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

A Hybrid Data-Driven and Model-Based Anomaly Detection Scheme for DER Operation: Preprint

This paper proposes a hybrid data and model-based anomaly detection for securing the operation of distributed energy resources (DERs) in distribution grids. Data-driven autoencoders (AE) are set up at the edge level by taking local DER data and detect anomalous operations by leveraging the reconstruction ability. In parallel, model-based state estimation (SE) is running at the system level by taking system models and measurements, the anomalies are identified by analyzing the measurements residual. The hybrid scheme preserves the benefits of both data-driven and model-based analysis and thus improves the robustness and accuracy of anomaly detection. It can be established by getting full use of the existing infrastructures in distribution grids. Numerical tests on a realistic distribution feeder in Southern California highlight the effectiveness as well as benefits of the proposed scheme.

anomaly detection↗

ENERGY SPECTRUM OF ENERGETIC PARTICLES ACCELERATED BY SHOCK WAVES: FROM FOCUSED TRANSPORT TO DIFFUSIVE ACCELERATION

The focused transport equation (FTE) includes all the necessary physics for modeling the shock acceleration of energetic particles with a unified description of first-order Fermi acceleration, shock drift acceleration, and shock surfing acceleration. It can treat the acceleration and transport of particles with an anisotropic distribution. In this study, the energy spectrum of pickup ions accelerated at shocks of various obliquities is investigated based on the FTE. We solve the FTE by using a stochastic approach. The shock acceleration leads to a two-component energy spectrum. The low-energy component of the spectrum is made up of particles that interact with shock one to a few times. For these particles, the pitch angle distribution is highly anisotropic, and the energy spectrum is variable depending on the momentum and pitch angle of injected particles. At high energies, the spectrum approaches a power law consistent with the standard diffusive shock acceleration (DSA) theory. For a parallel shock, the high-energy component of the power-law spectrum, with the spectral index being the same as the prediction of DSA theory, starts just a few times the injection speed. For an oblique or quasi-perpendicular shock, the high-energy component of the spectrum exhibits a double power-law distribution: a harder power-law spectrum followed by another power-law spectrum with a slope the same as the spectral index of DSA. The shock acceleration will eventually go into the DSA regime at higher energies even if the anisotropy is not small. The intensity of the energy spectrum given by the FTE, in the high-energy range where particles get efficient acceleration in the DSA regime, is different from that given by the standard DSA theory for the same injection source. We define the injection efficiency {eta} as the ratio between them. For a parallel shock, the injection efficiency is less than 1, but for an oblique shock or a quasi-perpendicular shock it could be greater.

79 ASTRONOMY AND ASTROPHYSICS↗

Stable parallel training of Wasserstein conditional generative adversarial neural networks

In this work, we propose a stable, parallel approach to train Wasserstein conditional generative adversarial neural networks (W-CGANs) under the constraint of a fixed computational budget. Differently from previous distributed GANs training techniques, our approach avoids inter-process communications, reduces the risk of mode collapse and enhances scalability by using multiple generators, each one of them concurrently trained on a single data label. The use of the Wasserstein metric also reduces the risk of cycling by stabilizing the training of each generator. We illustrate the approach on the CIFAR10, CIFAR100, and ImageNet1k datasets, three standard benchmark image datasets, maintaining the original resolution of the images for each dataset. Performance is assessed in terms of scalability and final accuracy within a limited fixed computational time and computational resources. To measure accuracy, we use the inception score, the Fréchet inception distance, and image quality. An improvement in inception score and Fréchet inception distance is shown in comparison to previous results obtained by performing the parallel approach on deep convolutional conditional generative adversarial neural networks as well as an improvement of image quality of the new images created by the GANs approach. Weak scaling is attained on both datasets using up to 2000 NVIDIA V100 GPUs on the OLCF supercomputer Summit.

97 MATHEMATICS AND COMPUTING↗

Dispersive Alfvén waves in a plasma with anisotropic superthermal particles

The dispersion of dispersive Alfvén wave in a low β plasma with anisotropic superthermal particles modeled by a bi-nonextensive distribution is derived from a kinetic way. The effect of anisotropic temperature on inertial Alfvén wave is so small that it is negligible. However, it will play an important role on the property of kinetic Alfvén wave (KAW). The numerical results reveal that the presence of superthermal electrons in the small wavenumber limit will lead the damping rate of the KAW bigger than the one with Maxwellian distribution. Whereas, the damping rate of KAW in the large wavenumber limit will decrease with the presence of superthermal electrons. When the effect of electron anisotropic temperature overwhelms the effect of finite ion gyroradius in the small wavenumber regime, the damping rate of KAW grows with the presence of electron temperature anisotropy. On the other hand, when the effects of finite ion gyroradius play a dominant role in the large wavenumber regime, the damping rate of KAW increases with the effective perpendicular and parallel electron temperatures.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Equilibrium and stability in a heliotron with anisotropic hot particle slowing-down distribution

The equilibrium and linear fluid Magnetohydrodynamic (MHD) stability in an inward-shifted large helical device heliotron configuration are investigated with the 3D ANIMEC and TERPSICHORE codes, respectively. A modified slowing-down distribution function is invoked to study anisotropic pressure conditions. An appropriate choice of coefficients and exponents allows the simulation of neutral beam injection in which the angle of injection is varied from parallel to perpendicular. The fluid stability analysis concentrates on the application of the Johnson-Kulsrud-Weimer energy principle. The growth rates are maximum at <{beta}>{approx}2%, decrease significantly at <{beta}>{approx}4.5%, do not vary significantly with variations of the injection angle and are similar to those predicted with a bi-Maxwellian hot particle distribution function model. Stability is predicted at <{beta}>{approx}2.5% with a sufficiently peaked energetic particle pressure profile. Electrostatic potential forms from the MHD instability necessary for guiding centre orbit following are calculated.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Angularly resolved photoionization dynamics in atoms and molecules combining temporally and spectrally resolved experiments at ATTOLab and Synchrotron SOLEIL

We report results for XUV-IR two-photon ionization of Ar, Ne, NO, and O2, where an XUV attosecond pulse train is superimposed with a synchronized IR pulse, obtained at the ATTOLab laser facility using electron–ion coincidence 3D momentum spectroscopy. Temporally resolved photoelectron angular distributions providing angle-resolved time-delays for np ionization of Ar and Ne, achieved by reconstruction of attosecond beating by interference of two-photon transitions through a unified formalism (Joseph et al. in J Phys B At Mol Opt Phys 53:184007, 2020), are summarized. For inner valence XUV-IR dissociative photoionization of NO and O2 molecules, we report electron–ion kinetic energy correlation diagrams and disentangle the dissociative photoionization processes relying on parallel XUV experiments at Synchrotron SOLEIL. For ionization into the NO+(c3Π) ionic state, extending the formalism developed for single-photon ionization, we focus on photoelectron angular distributions averaged on the delay between the XUV and the IR field in the field frame, molecular frame, and electron frame of reference.

Joseph, J↗