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 1,351 records · Page 75

Ion acceleration in Ar-Xe and Ar-He plasmas. II. Ion velocity distribution functions

Ion velocity distribution functions (ivdfs) are investigated by laser induced fluorescence in Ar-Xe and Ar-He expanding helicon plasmas as a function of gas composition. In the case of Ar-Xe plasma, it was found that in the helicon source, both the Ar{sup +} and Xe{sup +} vdfs are unimodal. Their parallel speeds are subsonic and unaffected by changes in gas composition. At the end of the source, the argon ivdf shows a bimodal structure indicative of an electric double layer upstream of the measurement location. The fast argon ion component parallel velocity increases with Xe fraction from 6.7 to 8 km/s as the Xe fraction increases from 0% to 4%. In the expansion region, the bimodal character of Ar ivdf is maintained with a supersonic fast component reaching parallel speeds of 10.5 km/s. For all the studied plasma conditions and different spatial locations, the Xe{sup +} vdf exhibits a unimodal structure with a maximum parallel flow velocity of 2.2 km/s at the end of the source. For Ar-He plasma, the Ar ivdf is bimodal with the fast ion component parallel velocity increasing from 5.2 to 7.8 km/s as the He fraction increases from 0% to 30%. For the same He fraction range, the slow argon ion population distribution changes from a single Gaussian to a wide distribution extending all the way from the speed of the fast population to 0 m/s.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

PEBBLES Mechanics Simulation Speedup

Pebble bed reactors contain large numbers of spherical fuel elements arranged randomly. Determining the motion and location of these fuel elements is required for calculating certain parameters of pebble bed reactor operation. These simulations involve hundreds of thousands of pebbles and involve determining the entire core motion as pebbles are recirculated. Single processor algorithms for this are insufficient since they would take decades to centuries of wall-clock time. This paper describes the process of parallelizing and speeding up the PEBBLES pebble mechanics simulation code. Both shared memory programming with the Open Multi-Processing API and distributed memory programming with the Message Passing Interface API are used in simultaneously in this process. A new shared memory lock-less linear time collision detection algorithm is described. This method allows faster detection of pebbles in contact than generic methods. These combine to make full recirculations on AVR sized reactors possible in months of wall clock time.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Distributed X-ray photon correlation spectroscopy data reduction using Hadoop MapReduce

Multi-speckle X-ray photon correlation spectroscopy (XPCS) is a powerful technique for characterizing the dynamic nature of complex materials over a range of time scales. XPCS has been successfully applied to study a wide range of systems. Recent developments in higher-frame-rate detectors, while aiding in the study of faster dynamical processes, creates large amounts of data that require parallel computational techniques to process in near real-time. Here, an implementation of the multi-tau and two-time autocorrelation algorithms using the Hadoop MapReduce framework for distributed computing is presented. In conclusion, the system scales well with regard to the increase in the data size, and has been serving the users of beamline 8-ID-I at the Advanced Photon Source for near real-time autocorrelations for the past five years.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Density and Magnetic Field Asymmetric Kelvin‐Helmholtz Instability

Abstract The Kelvin‐Helmholtz (KH) instability can transport mass, momentum, magnetic flux, and energy between the magnetosheath and magnetosphere, which plays an important role in the solar‐wind‐magnetosphere coupling process for different planets. Meanwhile, strong density and magnetic field asymmetry are often present between the magnetosheath (MSH) and magnetosphere (MSP), which could affect the transport processes driven by the KH instability. Our magnetohydrodynamics simulation shows that the KH growth rate is insensitive to the density ratio between the MSP and the MSH in the compressible regime, which is different than the prediction from linear incompressible theory. When the interplanetary magnetic field (IMF) is parallel to the planet's magnetic field, the nonlinear KH instability can drive a double mid‐latitude reconnection (DMLR) process. The total double reconnected flux depends on the KH wavelength and the strength of the lower magnetic field. When the IMF is anti‐parallel to the planet's magnetic field, the nonlinear interaction between magnetic reconnection and the KH instability leads to fast reconnection (i.e., close to Petschek reconnection even without including kinetic physics). However, the peak value of the reconnection rate still follows the asymmetric reconnection scaling laws. We also demonstrate that the DMLR process driven by the KH instability mixes the plasma from different regions and consequently generates different types of velocity distribution functions. We show that the counter‐streaming beams can be simply generated via the change of the flux tube connection and do not require parallel electric fields.

Astronomy & Astrophysics↗

Parallel-vector algorithms for particle simulations on shared-memory multiprocessors

Over the last few decades, the computational demands of massive particle-based simulations for both scientific and industrial purposes have been continuously increasing. Hence, considerable efforts are being made to develop parallel computing techniques on various platforms. In such simulations, particles freely move within a given space, and so on a distributed-memory system, load balancing, i.e., assigning an equal number of particles to each processor, is not guaranteed. However, shared-memory systems achieve better load balancing for particle models, but suffer from the intrinsic drawback of memory access competition, particularly during (1) paring of contact candidates from among neighboring particles and (2) force summation for each particle. Here, novel algorithms are proposed to overcome these two problems. For the first problem, the key is a pre-conditioning process during which particle labels are sorted by a cell label in the domain to which the particles belong. Then, a list of contact candidates is constructed by pairing the sorted particle labels. For the latter problem, a table comprising the list indexes of the contact candidate pairs is created and used to sum the contact forces acting on each particle for all contacts according to Newton's third law. With just these methods, memory access competition is avoided without additional redundant procedures. The parallel efficiency and compatibility of these two algorithms were evaluated in discrete element method (DEM) simulations on four types of shared-memory parallel computers: a multicore multiprocessor computer, scalar supercomputer, vector supercomputer, and graphics processing unit. The computational efficiency of a DEM code was found to be drastically improved with our algorithms on all but the scalar supercomputer. Thus, the developed parallel algorithms are useful on shared-memory parallel computers with sufficient memory bandwidth.

97 MATHEMATICS AND COMPUTING↗

AC Resistance Reduction Using Orthogonal Air Gaps in High-Frequency Inductors: Preprint

This paper presents a relatively simple technique to reduce winding losses due to fringing fields in high-frequency inductors. In high-frequency power electronics, ac inductor winding losses are affected by skin and proximity effects, including uneven current distribution due to fringing magnetic fields around airgaps. It is well known how fringing effects can be mitigated using distributed airgaps, at the expense of non-standard core or winding geometry. In planar magnetics, the orthogonal-airgap approach proposed in this paper combines airgaps in core segments parallel with the windings with conventionally placed airgaps in segments perpendicular to the windings. The approach is developed using a 1D analytical framework and validated by 2D finite-element simulations. Analytical guidelines are presented to optimize the airgaps to achieve minimum ac resistance. As a case study, a planar inductor is designed for an 8~kW SiC-based buck converter operating at 250~kHz. It is shown how the orthogonal airgaps result in more than 45\% reduction in ac resistance and substantially reduced inductor losses compared to the design using standard airgaps. The results are verified by loss measurements on an experimental converter prototype.

24 POWER TRANSMISSION AND DISTRIBUTION↗

New regime of low ion collisionality in the neoclassical equilibrium of tokamak plasmas

Here, the neoclassical description of an axisymmetric toroidal plasma equilibrium is formulated for an unconventionally low ordering of the collisionality that suits realistic thermonuclear fusion conditions. This requires a drift-kinetic analysis to the second order of the ion Larmor radius, which yields a new contribution to the leading solution for the non-Maxwellian part of the ion distribution function if the equilibrium geometry is not up-down symmetric. An explicit geometrical factor weighs this second Larmor-radius order, low-collisionality effect that modifies the neoclassical ion parallel flow, and the ion contribution to the bootstrap current.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Project X Accelerator R&D Plan

Project X is a high intensity proton facility conceived to support a world-leading program in neutrino and flavor physics over the next two decades at Fermilab. Project X is an integral part of the Fermilab Roadmap as described in the Fermilab Steering Group Report. Project X is based on an 8 GeV superconducting H-linac, paired with the existing (but modified) Main Injector and Recycler Ring, to provide in excess of 2 MW of beam power throughout the energy range 60-120 GeV, simultaneous with at least 100 kW of beam power at 8 GeV. The linac utilizes technology in common with the ILC over the energy range 0.6-8.0 GeV. Beam current parameters can be made identical to ILC resulting in identical rf generation and distribution systems. This alignment of ILC and Project X technologies allows for a shared development effort. The initial 0.6 GeV of the linac draws heavily on technology developed by Argonne National Laboratory for a facility for rare isotope beams. It is anticipated that the exact configuration and operating parameters of the linac will be defined through the R&D program and will retain alignment with the ILC plan as it evolves over this period. Utilization of the Recycler Ring as an H{sup -} stripper and accumulator ring is the key element that provides the flexibility to operate the linac with the same beam parameters as the ILC. The linac operates at 5 Hz with a total of 5.6 x 10{sup 13} H{sup -} ions delivered per pulse. H{sup -} are stripped at injection into the Recycler in a manner that 'paints' the beam both transversely and longitudinally to reduce space charge forces. Following the 1 ms injection, the orbit moves off the stripping foil and circulates for 200 msec, awaiting the next injection. Following three such injections a total of 1.7 x 10{sup 14} protons are transferred in a single turn to the Main Injector. These protons are then accelerated to 120 GeV and fast extracted to a neutrino target. The Main Injector cycle takes 1.4 seconds, producing approximately 2.3 MW of beam power at 120 GeV. At lower proton energies Main Injector cycle times can be shorter, allowing a beam power above 2 MW in the range of proton energy between 60 GeV and 120 GeV. In parallel, because the loading of the Recycler only requires 0.6 seconds, up to four linac cycles are available for accumulation and distribution of 8 GeV protons from the Recycler. Total available 8 GeV beam power lies in the range of 100-200 kW, depending on the energy in the Main Injector. Primary modifications to the existing accelerator complex to support Project X include integration of an H{sup -} injection system, a new RF system, a new extraction system, and measures to mitigate electron cloud effects, in the Recycler Ring. The Main Injector would need a new RF system, measures to preserve beam stability through transition, and measures to mitigate electron cloud effects. Finally, substantial modifications to the existing NuMI target station will be required to support >2 MW operations. It is anticipated that Project X configured as described above would initially support high intensity neutrino beams to the NOvA experiment, in parallel with at least one new 8 GeV based flavor/rare decay experiment. Depending upon future directions flexibility is retained for delivering neutrinos toward the DUSEL site and/or protons into the Tevatron. The purpose of this document is to describe an R&D plan that would position the U.S. to initiate construction of Project X in the 2012 time frame, assuming a go ahead decision in roughly 2010. The organization of this document is as follows: (1) Goals - Describes goals of the R&D and preliminary design period (2008-2011). Included are design, technical development, project documentation, and organizational goals. These are described in the context of an overall set of performance goals for Project X. (2) R&D Plan Elements - Describes the essential technical elements of the plan, including major subsystem performance requirements, associated accelerator and technology issues, and the plans for addressing these issues. (3) R&D Plan - Describes how the plan elements are assembled into a time-ordered plan, defines the associated resources required to support this plan, and describes the alignment of activities undertaken within the Project X, SRF/ILC, and HINS programs. Also describes how the R&D plan will be organized and executed by the prospective participating institutions.

43 PARTICLE ACCELERATORS↗

Mesh-free data transfer algorithms for partitioned multiphysics problems: Conservation, accuracy, and parallelism

In this study we analyze and extend mesh-free algorithms for three-dimensional data transfer problems in partitioned multiphysics simulations. We first provide a direct comparison between a mesh-based weighted residual method using the common-refinement scheme and two mesh-free algorithms leveraging compactly supported radial basis functions: one using a spline interpolation and one using a moving least square reconstruction. Through the comparison we assess both the conservation and accuracy of the data transfer obtained from each of the methods. We do so for a varying set of geometries with and without curvature and sharp features and for functions with and without smoothness and with varying gradients. Our results show that the mesh-based and mesh-free algorithms are complementary with cases where each was demonstrated to perform better than the other. We then focus on the mesh-free methods by developing a set of algorithms to parallelize them based on sparse linear algebra techniques. This includes a discussion of fast parallel radius searching in point clouds and restructuring the interpolation algorithms to leverage data structures and linear algebra services designed for large distributed computing environments. The scalability of our new algorithms is demonstrated on a leadership class computing facility using a set of basic scaling studies. Finally, these scaling studies show that for problems with reasonable load balance, our new algorithms for both spline interpolation and moving least square reconstruction demonstrate both strong and weak scalability using more than 100,000 MPI processes with billions of degrees of freedom in the data transfer operation.

97 MATHEMATICS AND COMPUTING↗

FitCache: A Transparent Drop-In Framework for Multi-Tier Caching to Accelerate Distributed Deep Learning Workloads

Training in Deep learning (DL) remains highly compute- and data-intensive, with I/O becoming a critical bottleneck as models and datasets scale. Recent studies report that data loading can dominate training time, especially on large-scale HPC systems with shared parallel file systems (PFS). Existing caching approaches either rely on single-tier designs or require intrusive modifications to training pipelines, limiting their portability and effectiveness. In this work, we present FitCache, a transparent drop-in framework for multi-tier caching to accelerate distributed DL training by coordinating fast local memory (e.g., DRAM, Persistent Memory (PMem)) and NVMe as hierarchical caches atop PFS. Our design adapts to hardware diversity, i.e., if NVMe is missing, memory transparently acts as a caching tier, ensuring stable performance. FitCache transparently intercepts I/O requests and issues concurrent fetches across all tiers, returning data from the fastest responder without centralized metadata or static redirection paths. FitCache adapts to dynamic workloads and heterogeneous clusters while maintaining POSIX compatibility. Experiments on Frontier (2048 GPUs) and smaller research clusters show that FitCache reduces training time by up to 40% and per-batch I/O latency by up to 71.6% compared to Lustre Orion PFS, offering a drop-in solution for scalable DL training.

Hu, Guangxing [ORNL] (ORCID:0009000283203614)↗

Scalable Contour Tree Computation by Data Parallel Peak Pruning

As data sets grow to exascale, automated data analysis and visualization are increasingly important, to intermediate human understanding and to reduce demands on disk storage via in situ analysis. Trends in architecture of high performance computing systems necessitate analysis algorithms to make effective use of combinations of massively multicore and distributed systems. One of the principal analytic tools is the contour tree, which analyses relationships between contours to identify features of more than local importance. Unfortunately, the predominant algorithms for computing the contour tree are explicitly serial, and founded on serial metaphors, which has limited the scalability of this form of analysis. While there is some work on distributed contour tree computation, and separately on hybrid GPU-CPU computation, there is no efficient algorithm with strong formal guarantees on performance allied with fast practical performance. Furthermore, we report the first shared SMP algorithm for fully parallel contour tree computation, with formal guarantees of O(lg V lg t) parallel steps and O(V lg V) work for data with V samples and t contour tree supernodes, and implementations with more than 30× parallel speed up on both CPU using TBB and GPU using Thrust and up 70× speed up compared to the serial sweep and merge algorithm.

97 MATHEMATICS AND COMPUTING↗

Distributed caching for processing raw arrays

As applications continue to generate multi-dimensional data at exponentially increasing rates, fast analytics to extract meaningful results is becoming extremely important. The database community has developed array databases that alleviate this problem through a series of techniques. In-situ mechanisms provide direct access to raw data in the original format---without loading and partitioning. Parallel processing scales to the largest datasets. In-memory caching reduces latency when the same data are accessed across a workload of queries. However, we are not aware of any work on distributed caching of multi-dimensional raw arrays. In this paper, we introduce a distributed framework for cost-based caching of multi-dimensional arrays in native format. Given a set of files that contain portions of an array and an online query workload, the framework computes an effective caching plan in two stages. First, the plan identifies the cells to be cached locally from each of the input files by continuously refining an evolving R-tree index. In the second stage, an optimal assignment of cells to nodes that collocates dependent cells in order to minimize the overall data transfer is determined. We design cache eviction and placement heuristic algorithms that consider the historical query workload. A thorough experimental evaluation over two real datasets in three file formats confirms the superiority -- by as much as two orders of magnitude -- of the proposed framework over existing techniques in terms of cache overhead and workload execution time.

Zhao, Weijie↗

Ion energy distribution near a plasma meniscus with beam extraction for multi element focused ion beams

An earlier study of the axial ion energy distribution in the extraction region (plasma meniscus) of a compact microwave plasma ion source showed that the axial ion energy spread near the meniscus is small ({approx}5 eV) and comparable to that of a liquid metal ion source, making it a promising candidate for focused ion beam (FIB) applications [J. V. Mathew and S. Bhattacharjee, J. Appl. Phys. 105, 96101 (2009)]. In the present work we have investigated the radial ion energy distribution (IED) under the influence of beam extraction. Initially a single Einzel lens system has been used for beam extraction with potentials up to -6 kV for obtaining parallel beams. In situ measurements of IED with extraction voltages upto -5 kV indicates that beam extraction has a weak influence on the energy spread ({+-}0.5 eV) which is of significance from the point of view of FIB applications. It is found that by reducing the geometrical acceptance angle at the ion energy analyzer probe, close to unidirectional distribution can be obtained with a spread that is smaller by at least 1 eV.

43 PARTICLE ACCELERATORS↗

Dosimetric equivalence of nonstandard HDR brachytherapy catheter patterns

Purpose: To determine whether alternative high dose rate prostate brachytherapy catheter patterns can result in similar or improved dose distributions while providing better access and reducing trauma. Materials and Methods: Standard prostate cancer high dose rate brachytherapy uses a regular grid of parallel needle positions to guide the catheter insertion. This geometry does not easily allow the physician to avoid piercing the critical structures near the penile bulb nor does it provide position flexibility in the case of pubic arch interference. This study used CT datasets with 3 mm slice spacing from ten previously treated patients and digitized new catheters following three hypothetical catheter patterns: conical, bi-conical, and fireworks. The conical patterns were used to accommodate a robotic delivery using a single entry point. The bi-conical and fireworks patterns were specifically designed to avoid the critical structures near the penile bulb. For each catheter distribution, a plan was optimized with the inverse planning algorithm, IPSA, and compared with the plan used for treatment. Irrelevant of catheter geometry, a plan must fulfill the RTOG-0321 dose criteria for target dose coverage (V{sub 100}{sup Prostate}>90%) and organ-at-risk dose sparing (V{sub 75}{sup Bladder}<1 cc, V{sub 75}{sup Rectum}<1 cc, V{sub 125}{sup Urethra}<<1 cc). Results: The three nonstandard catheter patterns used 16 nonparallel, straight divergent catheters, with entry points in the perineum. Thirty plans from ten patients with prostate sizes ranging from 26 to 89 cc were optimized. All nonstandard patterns fulfilled the RTOG criteria when the clinical plan did. In some cases, the dose distribution was improved by better sparing the organs-at-risk. Conclusion: Alternative catheter patterns can provide the physician with additional ways to treat patients previously considered unsuited for brachytherapy treatment (pubic arch interference) and facilitate robotic guidance of catheter insertion. In addition, alternative catheter patterns may decrease toxicity by avoidance of the critical structures near the penile bulb while still fulfilling the RTOG criteria.

61 RADIATION PROTECTION AND DOSIMETRY↗

A Fault Location Algorithm for Two-End Series-Compensated Double-Circuit Transmission Lines Using the Distributed Parameter Line Model

A new fault location algorithm for two-end series-compensated double-circuit transmission lines utilizing unsynchronized two-terminal current phasors and local voltage phasors is presented in this paper. The distributed parameter line model is adopted to take into account the shunt capacitance of the lines. The mutual coupling between the parallel lines in the zero-sequence network is also considered. The boundary conditions under different fault types are used to derive the fault location formulation. The developed algorithm directly uses the local voltage phasors on the line side of series compensation (SC) and metal oxide varistor (MOV). However, when potential transformers are not installed on the line side of SC and MOVs for the local terminal, these measurements can be calculated from the local terminal bus voltage and currents by estimating the voltages across the SC and MOVs. MATLAB SimPowerSystems is used to generate cases under diverse fault conditions to evaluating accuracy. The simulation results show that the proposed algorithm is qualified for practical implementation.

current phasors, distributed parameter line model,↗

Coupled‐cluster singles, doubles and perturbative triples with density fitting approximation for massively parallel heterogeneous platforms

Abstract A high‐performance implementation of the coupled‐cluster singles, doubles, and perturbative triples [CCSD(T)] is developed in the Massively Parallel Quantum Chemistry program. Novel features include: (1) reduced memory requirements via a density‐fitting (DF) CCSD implementation utilizing distributed lazy evaluation for tensors with more than two unoccupied indices and (2) the ability to utilize efficiently many‐core nodes (Intel Xeon Phi) and heterogeneous nodes with multiple NVIDIA GPUs on each node. All data that are greater than quadratic in the system size are distributed among processes. Excellent strong scaling is observed on distributed‐memory computers equipped with conventional CPUs, Intel Xeon Phi processors, and heterogeneous nodes with multiple NVIDIA GPUs Canonical CCSD(T) energies can be evaluated for systems containing 200 electrons and 1000 basis functions in a few days using a small size commodity cluster, with even larger computations possible on leadership‐class computing resources.

Peng, Chong↗

Coarse-grained simulation of colloidal self-assembly, cation exchange, and rheology in Na/Ca smectite clay gels

Knowledge Gap: The aggregation of clay minerals—layered silicate nanoparticles—strongly impacts fluid flow, solute migration, and solid mechanics in soils, sediments, and sedimentary rocks. Experimental and computational characterization of clay aggregation is inhibited by the delicate water-mediated nature of clay colloidal interactions and by the range of spatial scales involved, from 1 nm thick platelets to flocs with dimensions up to micrometers or more. Simulations: Using a new coarse-grained molecular dynamics (CGMD) approach, we predicted the microstructure, dynamics, and rheology of hydrated smectite (more precisely, montmorillonite) clay gels containing up to 2,000 clay platelets on length scales up to 0.1 μm. Further, simulations investigated the impact of simulation time, platelet diameters (6 to 25nm), and the ratio of Na to Ca exchangeable cations on the assembly of tactoids (i.e., stacks of parallel clay platelets) and larger aggregates (i.e., assemblages of tactoids). We analyzed structural features including tactoid size and size distribution, basal spacing, counterion distribution in the electrical double layer, clay association modes, and the rheological properties of smectite gels. Findings: Our results demonstrate new potential to characterize and understand clay aggregation in dilute suspensions and gels on a scale of thousands of particles with explicit representation of counterion clouds and with accuracy approaching that of all-atom molecular dynamics (MD) simulations. For example, our simulations predict the strong impact of Na/Ca ratio on clay tactoid formation and the shear-thinning rheology of clay gels.

42 ENGINEERING↗