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 667 records · Page 37

Using Direct Laboratory Measurements of Electron Temperature Anisotropy to Identify the Heating Mechanism in Electron-Only Guide Field Magnetic Reconnection

Anisotropic electron heating during electron-only magnetic reconnection with a large guide magnetic field is directly measured in a laboratory plasma through in situ measurements of electron velocity distribution functions. Electron heating preferentially parallel to the magnetic field is localized to one separatrix, and anisotropies of 1.5 are measured. The mechanism for electron energization is identified as the parallel reconnection electric field because of the anisotropic nature of the heating and spatial localization. These characteristics are reproduced in a 2D particle-in-cell simulation and are also consistent with numerous magnetosheath observations. A measured increase in the perpendicular temperature along both separatrices is not reproduced by our 2D simulations. Here, this work has implications for energy partition studies in magnetosheath and laboratory reconnection.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing

The recent development of deep learning has been mostly focusing on Euclidean data, such as images, videos, audios, etc. However, most real-world information and relation are often expressed as graphs. To efficiently learn from graph data, graph convolutional networks (GCNs) emerge as a promising approach, showing advantages in several practical applications such as social network analysis, knowledge discovery, 3D modeling, motion capturing, etc. Real-world graphs are usually extremely large and imbalanced, posting significant performance demand and design challenges on the hardware dedicated for GCN inference. In this paper, we propose an architecture design called UW-GCN to accelerate graph convolutional network inference. To tackle the major performance bottleneck from workload imbalance, we propose dynamic neighborhood stealing and remote chunk shuffling techniques, relying on hardware flexibility to achieve hardware auto-tuning under negligible area or delay overhead. Specifically, UW-GCN is able to smartly profile the sparse graph pattern while continuously adjusting the workload distribution via routing reconfiguration among parallel processing elements (PEs). The ideal configuration is then reused in the remaining iterations. To the best of our knowledge, this is the first accelerator design particularly for GCN and the first work relying on hardware auto-tuning, which is normally based on software, to achieve near-optimal workload balance in processing sparse structures.

Geng, Tong↗

GridPACK Toolkit for Developing Power Grid Simulations on High Performance Computing Platforms

This paper describes the GridPACK™ framework, which is designed to help power grid engineers develop modeling software capable of running on todays high performance computers. The framework contains modules for setting up distributed power grid networks, assigning buses and branches with arbitrary behaviors to the network, creating distributed matrices and vectors, using parallel linear and non-linear solvers to solve algebraic equations, and mapping functionality to create matrices and vectors based on properties of the network. In addition, the framework contains additional functionality to support IO and to manage errors.

Palmer, Bruce J.↗

Open | SpeedShop: An Open Source Infrastructure for Parallel Performance Analysis

Over the last decades a large number of performance tools has been developed to analyze and optimize high performance applications. Their acceptance by end users, however, has been slow: each tool alone is often limited in scope and comes with widely varying interfaces and workflow constraints, requiring different changes in the often complex build and execution infrastructure of the target application. We started the Open | SpeedShop project about 3 years ago to overcome these limitations and provide efficient, easy to apply, and integrated performance analysis for parallel systems. Open | SpeedShop has two different faces: it provides an interoperable tool set covering the most common analysis steps as well as a comprehensive plugin infrastructure for building new tools. In both cases, the tools can be deployed to large scale parallel applications using DPCL/Dyninst for distributed binary instrumentation. Further, all tools developed within or on top of Open | SpeedShop are accessible through multiple fully equivalent interfaces including an easy-to-use GUI as well as an interactive command line interface reducing the usage threshold for those tools.

Schulz, Martin↗

Clock Agreement Among Parallel Supercomputer Nodes

This dataset presents measurements that quantify the clock synchronization time-agreement characteristics among several high performance computers including the current world's most powerful machine for open science, the U.S. Department of Energy's Titan machine sited at Oak Ridge National Laboratory. These ultra-fast machines derive much of their computational capability from extreme node counts (over 18000 nodes in the case of the Titan machine). Time-agreement is commonly utilized by parallel programming applications and tools, distributed programming application and tools, and system software. Our time-agreement measurements detail the degree of time variance between nodes and how that variance changes over time. The dataset includes empirical measurements and the accompanying spreadsheets.

97 MATHEMATICS AND COMPUTING↗

Libra: Scalable Load Balance Analysis

Libra is a tool for scalable analysis of load balance data from all processes in a parallel application. Libra contains an instrumentation module that collects model data from parallel applications and a parallel compression mechanism that uses distributed wavelet transforms to gather load balance model data in a scalable fashion. Data is output to files, and these files can be viewed in a GUI tool by Libra users. The GUI tool associates particular load balance data with regions for code, emabling users to view the load balance properties of distributed "slices" of their application code.

Source record↗

Formulas for robust, one-pass parallel computation of covariances and arbitrary-order statistical moments.

We present a formula for the pairwise update of arbitrary-order centered statistical moments. This formula is of particular interest to compute such moments in parallel for large-scale, distributed data sets. As a corollary, we indicate a specialization of this formula for incremental updates, of particular interest to streaming implementations. Finally, we provide pairwise and incremental update formulas for the covariance. Centered statistical moments are one of the most widely used tools in descriptive statistics. It is therefore essential for statistical analysis packages that robust and efficient algorithms be devised and implemented. However, robustness and speed of execution, in this context as well as in others, tend to be orthogonal. For instance, it is well known1 that algorithms for calculating centered statistical moments that utilize sum of powers for the sake of execution speed (one-pass algorithms) lead to unacceptable numerical instability.

99 GENERAL AND MISCELLANEOUS↗

Libra: Scalable Load Balance Analysis

Libra is a tool for scalable analysis of load balance data from all processes in a parallel application. Libra contains an instrumentation module that collects model data from parallel applications and a parallel compression mechanism that uses distributed wavelet transforms to gather load balance model data in a scalable fashion. Data is output to files, and these files can be viewed in a GUI tool by Libra users. The GUI tool associates particular load balance data with regions for code, emabling users to view the load balance properties of distributed "slices" of their application code.

Source record↗

Mechanical design of deformation compensated flexural pivots structured for linear nanopositioning stages

A method and deformation compensated flexural pivots structured for precision linear nanopositioning stages are provided. A deformation-compensated flexural linear guiding mechanism includes a basic parallel mechanism including a U-shaped member and a pair of parallel bars linked to respective pairs of I-link bars and each of the I-bars coupled by a respective pair of flexural pivots. The basic parallel mechanism includes substantially evenly distributed flexural pivots minimizing center shift dynamic errors.

42 ENGINEERING↗

Dynamic Phasor Modeling of Three Phase Voltage Source Inverters

With the increase in the development and implementation of distributed energy resources, application of parallel connected inverters is increasing and as a result having accurate modeling and simulation tools that can help in the design, analysis and stability assessment of the grid is of great importance. Development of fundamental methods that can achieve accurate, reliable, and computationally efficient results can be very beneficial. Application of Dynamic phasor (DP) modeling method has been limited to study of limited harmonics and small combination of interconnected converters due to the complexity associated with developing models that describe larger systems. In this paper, application of DP modeling method is expanded to model any number of parallel connected three phase voltage source inverter (VSI) with inclusion of a wider harmonic content including fundamental, subharmonics, inter-harmonics, switching frequency and their sidebands. Results achieved from this modeling method is compared with conventional average model as well as detailed switching model and the effect of inclusion of wider harmonic content on accuracy of DP modeling method is demonstrated.

Xue, Yaosuo↗

Improved parallel data partitioning by nested dissection with applications to information retrieval.

The computational work in many information retrieval and analysis algorithms is based on sparse linear algebra. Sparse matrix-vector multiplication is a common kernel in many of these computations. Thus, an important related combinatorial problem in parallel computing is how to distribute the matrix and the vectors among processors so as to minimize the communication cost. We focus on minimizing the total communication volume while keeping the computation balanced across processes. In [1], the first two authors presented a new 2D partitioning method, the nested dissection partitioning algorithm. In this paper, we improve on that algorithm and show that it is a good option for data partitioning in information retrieval. We also show partitioning time can be substantially reduced by using the SCOTCH software, and quality improves in some cases, too.

99 GENERAL AND MISCELLANEOUS↗

Work stealing for GPU-accelerated parallel programs in a global address space framework

Task parallelism is an attractive approach to automatically load balance the computation in a parallel system and adapt to dynamism exhibited by parallel systems. Exploiting task parallelism through work stealing has been extensively studied in shared and distributed-memory contexts. In this paper, we study the design of a system that uses work stealing for dynamic load balancing of task-parallel programs executed on hybrid distributed-memory CPU-graphics processing unit (GPU) systems in a global-address space framework. We take into account the unique nature of the accelerator model employed by GPUs, the significant performance difference between GPU and CPU execution as a function of problem size, and the distinct CPU and GPU memory domains. We consider various alternatives in designing a distributed work stealing algorithm for CPU-GPU systems, while taking into account the impact of task distribution and data movement overheads. These strategies are evaluated using microbenchmarks that capture various execution configurations as well as the state-of-the-art CCSD(T) application module from the computational chemistry domain

97 MATHEMATICS AND COMPUTING↗

PLASMA EFFECTS ON FAST PAIR BEAMS. II. REACTIVE VERSUS KINETIC INSTABILITY OF PARALLEL ELECTROSTATIC WAVES

The interaction of TeV gamma-rays from distant blazars with the extragalactic background light produces relativistic electron-positron pair beams by the photon-photon annihilation process. Using the linear instability analysis in the kinetic limit, which properly accounts for the longitudinal and the small but finite perpendicular momentum spread in the pair momentum distribution function, the growth rate of parallel propagating electrostatic oscillations in the intergalactic medium is calculated. Contrary to the claims of Miniati and Elyiv, we find that neither the longitudinal nor the perpendicular spread in the relativistic pair distribution function significantly affect the electrostatic growth rates. The maximum kinetic growth rate for no perpendicular spread is even about an order of magnitude greater than the corresponding reactive maximum growth rate. The reduction factors in the maximum growth rate due to the finite perpendicular spread in the pair distribution function are tiny and always less than 10{sup –4}. We confirm earlier conclusions by Broderick et al. and our group that the created pair beam distribution function is quickly unstable in the unmagnetized intergalactic medium. Therefore, there is no need to require the existence of small intergalactic magnetic fields to scatter the produced pairs, so that the explanation (made by several authors) for the Fermi non-detection of the inverse Compton scattered GeV gamma-rays by a finite deflecting intergalactic magnetic field is not necessary. In particular, the various derived lower bounds for the intergalactic magnetic fields are invalid due to the pair beam instability argument.

79 ASTRONOMY AND ASTROPHYSICS↗

Calculation of the piezoelectric and flexoelectric effects in nanowires using a decoupled finite element analysis method

A simple and effective decoupled finite element analysis method was developed for simulating both the piezoelectric and flexoelectric effects of zinc oxide (ZnO) and barium titanate (BTO) nanowires (NWs). The piezoelectric potential distribution on a ZnO NW was calculated under three deformation conditions (cantilever, three-point, and four-point bending) and compared to the conventional fully coupled method. The discrepancies of the electric potential maximums from these two methods were found very small, validating the accuracy and effectiveness of the decoupled method. Both ZnO and BTO NWs yielded very similar potential distributions. Comparing the potential distributions induced by the piezoelectric and flexoelectric effects, we identified that the middle segment of a four-point bending NW beam is the ideal place for measuring the flexoelectric coefficient, because the uniform parallel plate capacitor-like potential distribution in this region is exclusively induced by the flexoelectric effect. This decoupled method could provide a valuable guideline for experimental measurements of the piezoelectric effects and flexoelectric effects in the nanometer scale.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Ion acceleration and heating by kinetic Alfvén waves associated with magnetic reconnection

In a previous study on the generation and signatures of kinetic Alfv en waves (KAWs) associated with magnetic reconnection in a current sheet revealed that KAWs are a common feature during reconnection [Liang et al. J. Geophys. Res.: Space Phys. 121, 6526 (2016)]. In this paper, ion acceleration and heating by the KAWs generated during magnetic reconnection are investigated with a three-dimensional (3-D) hybrid model. It is found that in the outflow region, a fraction of inflow ions are accelerated by the KAWs generated in the leading bulge region of reconnection, and their parallel velocities gradually increase up to slightly super-Alfv enic. As a result of waveparticle interactions, an accelerated ion beam forms in the direction of the anti-parallel magnetic field, in addition to the core ion population, leading to the development of non-Maxwellian velocity distributions, which include a trapped population with parallel velocities consistent with the wave speed. We then heat ions in both parallel and perpendicular directions. In the parallel direction, the heating results from nonlinear Landau resonance of trapped ions. In the perpendicular direction, however, evidence of stochastic heating by the KAWs is found during the acceleration stage, with an increase of magnetic moment μ. The coherence in the T $\perp$ ion temperature and the perpendicular electric and magnetic fields of KAWs also provides evidence for perpendicular heating by KAWs. The parallel and perpendicular heating of the accelerated beam occur simultaneously, leading to the development of temperature anisotropy with the perpendicular temperature T $\perp$ >T $\parallel$ temperature. The heating rate agrees with the damping rate of the KAWs, and the heating is dominated by the accelerated ion beam. In the later stage, with the increase of the fraction of the accelerated ions, interaction between the accelerated beam and the core population also contributes to the ion heating, ultimately leading to overlap of the beams and an overall anisotropy with T $\perp$ >T $\parallel$ .

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Effect on the divertor and scrape-off layer plasma properties of the distribution of power and particle influxes from the core

Calculations of the profiles along the field lines within the divertor and scrape-off layer (SOL) of differences in the plasma ion density, temperature, parallel current, parallel flow velocity, and electrostatic potential, which result from using different poloidal distributions of the particle and heat influxes crossing the separatrix from the core plasma into the SOL, are presented and discussed vis-a-vis experimental observations. The calculations show that the stronger outboard than inboard particle and heat fluxes into the SOL caused by the geometric compression/expansion of flux surfaces predicted by magnetohydrodynamic equilibrium calculations lead to a prediction of higher plasma temperature at the outboard divertor than at the inboard divertor, a result that is consistent with experimental observation and that confirms a previous prediction (made without accounting for drifts) of a possible cause of the observed in-out divertor power asymmetry. The calculations also illustrate the effect of the poloidal distribution of particle and power influx into the SOL on the flow velocity, parallel current, and electrostatic potential distributions in the SOL and divertor.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Generation of feasible integer solutions on a massively parallel computer using the feasibility pump

Here, we present an approach to parallelize generation of feasible mixed integer solutions of mixed integerlinear programs in distributed memory high performance computing environments. This approachcombines a parallel framework with feasibility pump (FP) as the rounding heuristic. It runs multiple FPinstances with different starting solutions concurrently, while allowing them to share information. Ourcomputational results suggest that the improvement resulting from parallelization using our approach is statistically significant.

97 MATHEMATICS AND COMPUTING↗

MerAligner: A Fully Parallel Sequence Aligner

Aligning a set of query sequences to a set of target sequences is an important task in bioinformatics. In this work we present merAligner, a highly parallel sequence aligner that implements a seed-and-extend algorithm and employs parallelism in all of its components. MerAligner relies on a high performance distributed hash table (seed index) and uses onesided communication capabilities of the Unified Parallel C to facilitate a fine-grained parallelism. We leverage communication optimizations at the construction of the distributed hash table and software caching schemes to reduce communication during the aligning phase. Additionally, merAligner preprocesses the target sequences to extract properties enabling exact sequence matching with minimal communication. Finally, we efficiently parallelize the I/O intensive phases and implement an effective load balancing scheme. Results show that merAligner exhibits efficient scaling up to thousands of cores on a Cray XC30 supercomputer using real human and wheat genome data while significantly outperforming existing parallel alignment tools.

Georganas, E↗