Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Graphics processing units”

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 361 records · Page 20
All sourcesNASA NTRSDOE OSTI
DOE OSTI2014

How empty are disk gaps opened by giant planets?

Gap clearing by giant planets has been proposed to explain the optically thin cavities observed in many protoplanetary disks. How much material remains in the gap determines not only how detectable young planets are in their birth environments, but also how strong co-rotation torques are, which impacts how planets can survive fast orbital migration. We determine numerically how the average surface density inside the gap, Σ{sub gap}, depends on planet-to-star mass ratio q, Shakura-Sunyaev viscosity parameter α, and disk height-to-radius aspect ratio h/r. Our results are derived from our new graphics processing unit accelerated Lagrangian hydrodynamical code PEnGUIn and are verified by independent simulations with ZEUS90. For Jupiter-like planets, we find Σ{sub gap}∝q {sup –2.2}α{sup 1.4}(h/r){sup 6.6}, and for near brown dwarf masses, Σ{sub gap}∝q {sup –1}α{sup 1.3}(h/r){sup 6.1}. Surface density contrasts inside and outside gaps can be as large as 10{sup 4}, even when the planet does not accrete. We derive a simple analytic scaling, Σ{sub gap}∝q {sup –2}α{sup 1}(h/r){sup 5}, that compares reasonably well to empirical results, especially at low Neptune-like masses, and use discrepancies to highlight areas for progress.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2014

Hunting a wandering supermassive black hole in the M31 halo hermitage

In the hierarchical structure formation scenario, galaxies enlarge through multiple merging events with less massive galaxies. In addition, the Magorrian relation indicates that almost all galaxies are occupied by a central supermassive black hole (SMBH) of mass 10{sup –3} times the mass of its spheroidal component. Consequently, SMBHs are expected to wander in the halos of their host galaxies following a galaxy collision, although evidence of this activity is currently lacking. We investigate a current plausible location of an SMBH wandering in the halo of the Andromeda galaxy (M31). According to theoretical studies of N-body simulations, some of the many substructures in the M31 halo are remnants of a minor merger occurring about 1 Gyr ago. First, to evaluate the possible parameter space of the infalling orbit of the progenitor, we perform numerous parameter studies using a graphics processing unit cluster. To reduce uncertainties in the predicted position of the expected SMBH, we then calculate the time evolution of the SMBH in the progenitor dwarf galaxy from N-body simulations using the plausible parameter sets. Our results show that the SMBH lies within the halo (∼20-50 kpc from the M31 center), closer to the Milky Way than the M31 disk. Furthermore, the predicted current positions of the SMBH were restricted to an observational field of 0.°6 × 0.°7 in the northeast region of the M31 halo. We also discuss the origin of the infalling orbit of the satellite galaxy and its relationships with the recently discovered vast thin disk plane of satellite galaxies around M31.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2014

Expansion techniques for collisionless stellar dynamical simulations

We present graphics processing unit (GPU) implementations of two fast force calculation methods based on series expansions of the Poisson equation. One method is the self-consistent field (SCF) method, which is a Fourier-like expansion of the density field in some basis set; the other method is the multipole expansion (MEX) method, which is a Taylor-like expansion of the Green's function. MEX, which has been advocated in the past, has not gained as much popularity as SCF. Both are particle-field methods and optimized for collisionless galactic dynamics, but while SCF is a 'pure' expansion, MEX is an expansion in just the angular part; thus, MEX is capable of capturing radial structure easily, while SCF needs a large number of radial terms. We show that despite the expansion bias, these methods are more accurate than direct techniques for the same number of particles. The performance of our GPU code, which we call ETICS, is profiled and compared to a CPU implementation. On the tested GPU hardware, a full force calculation for one million particles took ∼0.1 s (depending on expansion cutoff), making simulations with as many as 10{sup 8} particles fast for a comparatively small number of nodes.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

THE POWER OF IMAGING: CONSTRAINING THE PLASMA PROPERTIES OF GRMHD SIMULATIONS USING EHT OBSERVATIONS OF Sgr A*

Recent advances in general relativistic magnetohydrodynamic simulations have expanded and improved our understanding of the dynamics of black-hole accretion disks. However, current simulations do not capture the thermodynamics of electrons in the low density accreting plasma. This poses a significant challenge in predicting accretion flow images and spectra from first principles. Because of this, simplified emission models have often been used, with widely different configurations (e.g., disk- versus jet-dominated emission), and were able to account for the observed spectral properties of accreting black holes. Exploring the large parameter space introduced by such models, however, requires significant computational power that exceeds conventional computational facilities. In this paper, we use GRay, a fast graphics processing unit (GPU) based ray-tracing algorithm, on the GPU cluster El Gato, to compute images and spectra for a set of six general relativistic magnetohydrodynamic simulations with different magnetic field configurations and black-hole spins. We also employ two different parametric models for the plasma thermodynamics in each of the simulations. We show that, if only the spectral properties of Sgr A* are used, all 12 models tested here can fit the spectra equally well. However, when combined with the measurement of the image size of the emission using the Event Horizon Telescope, current observations rule out all models with strong funnel emission, because the funnels are typically very extended. Our study shows that images of accretion flows with horizon-scale resolution offer a powerful tool in understanding accretion flows around black holes and their thermodynamic properties.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

HELIOS-K: An ultrafast, open-source opacity calculator for radiative transfer

We present an ultrafast opacity calculator that we name HELIOS-K. It takes a line list as an input, computes the shape of each spectral line, and provides an option for grouping an enormous number of lines into a manageable number of bins. We implement a combination of Algorithm 916 and Gauss–Hermite quadrature to compute the Voigt profile, write the code in CUDA, and optimize the computation for graphics processing units (GPUs). We restate the theory of the k-distribution method and use it to reduce ∼10{sup 5}–10{sup 8} lines to ∼10–10{sup 4} wavenumber bins, which may then be used for radiative transfer, atmospheric retrieval and general circulation models. The choice of line-wing cutoff for the Voigt profile is a significant source of error and affects the value of the computed flux by ∼10%. This is an outstanding physical (rather than computational) problem, due to our incomplete knowledge of pressure broadening of spectral lines in the far line wings. We emphasize that this problem remains regardless of whether one performs line-by-line calculations or uses the k-distribution method and affects all calculations of exoplanetary atmospheres requiring the use of wavelength-dependent opacities. We elucidate the correlated-k approximation and demonstrate that it applies equally to inhomogeneous atmospheres with a single atomic/molecular species or homogeneous atmospheres with multiple species. Using a NVIDIA K20 GPU, HELIOS-K is capable of computing an opacity function with ∼10{sup 5} spectral lines in ∼1 s and is publicly available as part of the Exoclimes Simulation Platform (www.exoclime.org).

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

The 3D flow field around an embedded planet

3D modifications to the well-studied 2D flow topology around an embedded planet have the potential to resolve long-standing problems in planet formation theory. We present a detailed analysis of the 3D isothermal flow field around a 5 Earth-mass planet on a fixed circular orbit, simulated using our graphics processing unit hydrodynamics code PEnGUIn. We find that, overall, the horseshoe region has a columnar structure extending vertically much beyond the Hill sphere of the planet. This columnar structure is only broken for some of the widest horseshoe streamlines, along which high altitude fluid descends rapidly into the planet’s Bondi sphere, performs one horseshoe turn, and exits the Bondi sphere radially in the midplane. A portion of this flow exits the horseshoe region altogether, which we refer to as the “transient” horseshoe flow. The flow continues as it rolls up into a pair of up–down symmetric horizontal vortex lines shed into the wake of the planet. This flow, unique to 3D, affects both planet accretion and migration. It prevents the planet from sustaining a hydrostatic atmosphere due to its intrusion into the Bondi sphere, and leads to a significant corotation torque on the planet, unanticipated by 2D analysis. In the reported simulation, starting with a Σ∼r{sup −3/2} radial surface density profile, this torque is positive and partially cancels with the negative differential Lindblad torque, resulting in a factor of three slower planet migration rate. Finally, we report 3D effects can be suppressed by a sufficiently large disk viscosity, leading to results similar to 2D.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

CHOLLA: A NEW MASSIVELY PARALLEL HYDRODYNAMICS CODE FOR ASTROPHYSICAL SIMULATION

We present Computational Hydrodynamics On ParaLLel Architectures (Cholla ), a new three-dimensional hydrodynamics code that harnesses the power of graphics processing units (GPUs) to accelerate astrophysical simulations. Cholla models the Euler equations on a static mesh using state-of-the-art techniques, including the unsplit Corner Transport Upwind algorithm, a variety of exact and approximate Riemann solvers, and multiple spatial reconstruction techniques including the piecewise parabolic method (PPM). Using GPUs, Cholla evolves the fluid properties of thousands of cells simultaneously and can update over 10 million cells per GPU-second while using an exact Riemann solver and PPM reconstruction. Owing to the massively parallel architecture of GPUs and the design of the Cholla code, astrophysical simulations with physically interesting grid resolutions (≳256{sup 3}) can easily be computed on a single device. We use the Message Passing Interface library to extend calculations onto multiple devices and demonstrate nearly ideal scaling beyond 64 GPUs. A suite of test problems highlights the physical accuracy of our modeling and provides a useful comparison to other codes. We then use Cholla to simulate the interaction of a shock wave with a gas cloud in the interstellar medium, showing that the evolution of the cloud is highly dependent on its density structure. We reconcile the computed mixing time of a turbulent cloud with a realistic density distribution destroyed by a strong shock with the existing analytic theory for spherical cloud destruction by describing the system in terms of its median gas density.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2019

GPU-accelerated dislocation dynamics using subcycling time-integration

Discrete dislocation dynamics (DDD) simulations are being increasingly employed to investigate metal plasticity at the mesoscale. However, in spite of its ability to access beyond the length and time limits of atomistic methods, the DDD model is still limited by its high computational cost, with ranges of achievable strains too low and strain rates too high by several orders of magnitude compared with typical experimental conditions. By combining the efficiency of the recently developed subcycling time-integrator with the highly parallel architecture of graphics processing unit (GPU) devices, we developed a DDD model that provides significant acceleration compared to existing implementations. Our GPU-accelerated implementation enables large-scale DDD simulations that can reach relevant levels of strain using a moderate amount of computational resources.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Near real-time streaming analysis of big fusion data

Experiments on fusion plasmas produce high-dimensional data time series with ever-increasing magnitude and velocity, but turn-around times for analysis of this data have not kept up. For example, many data analysis tasks are often performed in a manual, ad-hoc manner some time after an experiment. In this article, we introduce the Delta framework that facilitates near real-time streaming analysis of big and fast fusion data. By streaming measurement data from fusion experiments to a high-performance compute center, Delta allows computationally expensive data analysis tasks to be performed in between plasma pulses. This article describes the modular and expandable software architecture of Delta and presents performance benchmarks of individual components as well as of an example workflow. Focusing on a streaming analysis workflow where electron cyclotron emission imaging (ECEi) data is measured at KSTAR on the National Energy Research Scientific Computing Center's (NERSC's) supercomputer we routinely observe data transfer rates of about 4 Gigabit per second. In NERSC, a demanding turbulence analysis workflow effectively utilizes multiple nodes and graphical processing units and executes them in under 5 min. We further discuss how Delta uses modern database systems and container orchestration services to provide web-based real-time data visualization. For the case of ECEi data we demonstrate how data visualizations can be augmented with outputs from machine learning models. Here, by providing session leaders and physics operators, results of higher-order data analysis using live visualizations may make more informed decisions on how to configure the machine for the next shot.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

Harmonic analysis of discrete tracers of large-scale structure

It is commonplace in cosmology to analyze fields projected onto the celestial sphere, and in particular density fields that are defined by a set of points e.g. galaxies. When performing an harmonic-space analysis of such data (e.g. an angular power spectrum) using a pixelized map one has to deal with aliasing of small-scale power and pixel window functions. We compare and contrast the approaches to this problem taken in the cosmic microwave background and large-scale structure communities, and advocate for a direct approach that avoids pixelization. We describe a method for performing a pseudo-spectrum analysis of a galaxy data set and show that it can be implemented efficiently using well-known algorithms for special functions that are suited to acceleration by graphics processing units (GPUs). The method returns the same spectra as the more traditional map-based approach if in the latter the number of pixels is taken to be sufficiently large and the mask is well sampled. The method is readily generalizable to cross-spectra and higher-order functions. It also provides a convenient route for distributing the information in a galaxy catalog directly in harmonic space, as a complement to releasing the configuration-space positions and weights, and a route to spectral apodization. Finally, we make public a code enabling the application of our method to existing and upcoming datasets.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Organizing Large Data Sets for Efficient Analyses on HPC Systems

Upcoming exascale applications could introduce significant data management challenges due to their large sizes, dynamic work distribution, and involvement of accelerators such as graphical processing units, GPUs. In this work, we explore the performance of reading and writing operations involving one such scientific application on two different supercomputers. Our tests showed that the Adaptable Input and Output System, ADIOS, was able to achieve speeds over 1TB/s, a significant fraction of the peak I/O performance on Summit. We also demonstrated the querying functionality in ADIOS could effectively support common selective data analysis operations, such as conditional histograms. In tests, this query mechanism was able to reduce the execution time by a factor of five. More importantly, ADIOS data management framework allows us to achieve these performance improvements with only a minimal amount of coding effort.

Gu, Junmin↗
DOE OSTI2023

Evaluation of Portable Programming Models to Accelerate LArTPC Detector Simulations

The Liquid Argon Time Projection Chamber (LArTPC) technology is widely used in high energy physics experiments, including the upcoming Deep Underground Neutrino Experiment (DUNE). Accurately simulating LArTPC detector responses is essential for analysis algorithm development and physics model interpretations. Accurate LArTPC detector response simulations are computationally demanding, and can become a bottleneck in the analysis workflow. Compute devices such as General-Purpose Graphics Processing Units (GPGPUs) have the potential to substantially accelerate simulations compared to traditional CPU-only processing. The software development for these compute accelerators often carries the cost of specialized code refactorization and porting to match the target hardware architecture. With the rapid evolution and increased diversity of the computer architecture landscape, it is highly desirable to have a portable solution that also maintains reasonable performance. We report our ongoing effort in evaluating Kokkos as a basis for this portable programming model using LArTPC simulations in the context of the Wire-Cell Toolkit, a C++ library for LArTPC simulations, data analysis, reconstruction and visualization.

47 OTHER INSTRUMENTATION↗
DOE OSTI2013

Many-core applications to online track reconstruction in HEP experiments

Interest in parallel architectures applied to real time selections is growing in High Energy Physics (HEP) experiments. In this paper we describe performance measurements of Graphic Processing Units (GPUs) and Intel Many Integrated Core architecture (MIC) when applied to a typical HEP online task: the selection of events based on the trajectories of charged particles. We use as benchmark a scaled-up version of the algorithm used at CDF experiment at Tevatron for online track reconstruction – the SVT algorithm – as a realistic test-case for low-latency trigger systems using new computing architectures for LHC experiment. We examine the complexity/performance trade-off in porting existing serial algorithms to many-core devices. Measurements of both data processing and data transfer latency are shown, considering different I/O strategies to/from the parallel devices.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2014

High Energy Electromagnetic Particle Transportation on the GPU

We present massively parallel high energy electromagnetic particle transportation through a finely segmented detector on a Graphics Processing Unit (GPU). Simulating events of energetic particle decay in a general-purpose high energy physics (HEP) detector requires intensive computing resources, due to the complexity of the geometry as well as physics processes applied to particles copiously produced by primary collisions and secondary interactions. The recent advent of hardware architectures of many-core or accelerated processors provides the variety of concurrent programming models applicable not only for the high performance parallel computing, but also for the conventional computing intensive application such as the HEP detector simulation. The components of our prototype are a transportation process under a non-uniform magnetic field, geometry navigation with a set of solid shapes and materials, electromagnetic physics processes for electrons and photons, and an interface to a framework that dispatches bundles of tracks in a highly vectorized manner optimizing for spatial locality and throughput. Core algorithms and methods are excerpted from the Geant4 toolkit, and are modified and optimized for the GPU application. Program kernels written in C/C++ are designed to be compatible with CUDA and OpenCL and with the aim to be generic enough for easy porting to future programming models and hardware architectures. To improve throughput by overlapping data transfers with kernel execution, multiple CUDA streams are used. Issues with floating point accuracy, random numbers generation, data structure, kernel divergences and register spills are also considered. Performance evaluation for the relative speedup compared to the corresponding sequential execution on CPU is presented as well.

Canal, P.↗
DOE OSTI2025

Track reconstruction as a service for collider physics

Optimizing charged-particle track reconstruction algorithms is crucial for efficient event reconstruction in Large Hadron Collider (LHC) experiments due to their significant computational demands. Existing track reconstruction algorithms have been adapted to run on massively parallel coprocessors, such as graphics processing units (GPUs), to reduce processing time. Nevertheless, challenges remain in fully harnessing the computational capacity of coprocessors in a scalable and non-disruptive manner. This paper proposes an inference-as-a-service approach for particle tracking in high energy physics experiments. To evaluate the efficacy of this approach, two distinct tracking algorithms are tested: Patatrack, a rule-based algorithm, and Exa.TrkX, a machine learning-based algorithm. The as-a-service implementations show enhanced GPU utilization and can process requests from multiple CPU cores concurrently without increasing per-request latency. The impact of data transfer is minimal and insignificant compared to running on local coprocessors. This approach greatly improves the computational efficiency of charged particle tracking, providing a solution to the computing challenges anticipated in the High-Luminosity LHC era.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2021

An efficient numerical model for predicting residual stress and strain in parts manufactured by laser powder bed fusion

Abstract Computational modeling of additively manufactured structures plays an increasingly important role in product design and optimization. For laser powder bed fusion processes, the accurate modeling of stress and distortion requires large amount of computational cost due to very localized heat input and evolving complex geometries. The current study takes advantage of a graphics processing unit accelerated explicit finite element analysis code and approximated heat conduction analysis to predict the macroscopic thermo-mechanical behavior in laser selective melting. Adjacent layers and tracks were lumped to reduce the number of time steps and elements in the finite element model. The effects of track and layer grouping on prediction accuracy and solution efficiency are investigated to provide a guidance for a cost-effective simulation. Thin-wall builds from Inconel alloy 625 (IN625) powders were simulated by applying the developed modeling approach to get the detailed residual stress and distortion at a computational speed 50 times higher than conventional approach. Under repeated heating and cooling cycles, a high tensile stress was produced near surfaces of a build due to a larger shrinkage on surface than that in central area. It is also shown that horizontal stresses concentrate near the root and top layers of the IN625 build. The predicted residual elastic strain distribution was validated by the experimental measurement using x-ray synchrotron diffraction.

36 MATERIALS SCIENCE↗
DOE OSTI2021

GPU coprocessors as a service for deep learning inference in high energy physics

In the next decade, the demands for computing in large scientific experiments are expected to grow tremendously. During the same time period, CPU performance increases will be limited. At the CERN Large Hadron Collider (LHC), these two issues will confront one another as the collider is upgraded for high luminosity running. Alternative processors such as graphics processing units (GPUs) can resolve this confrontation provided that algorithms can be sufficiently accelerated. In many cases, algorithmic speedups are found to be largest through the adoption of deep learning algorithms. We present a comprehensive exploration of the use of GPU-based hardware acceleration for deep learning inference within the data reconstruction workflow of high energy physics. We present several realistic examples and discuss a strategy for the seamless integration of coprocessors so that the LHC can maintain, if not exceed, its current performance throughout its running.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Geometric GNNs for charged particle tracking at GlueX

Nuclear physics experiments are aimed at uncovering the fundamental building blocks of matter. The experiments involve high-energy collisions that produce complex events with many particle trajectories. Tracking charged particles resulting from collisions in the presence of a strong magnetic field is critical to enable the reconstruction of particle trajectories and precise determination of interactions. It is traditionally achieved through combinatorial approaches that scale worse than linearly as the number of hits grows. Since particle hit data naturally form a point cloud and can be structured as graphs, graph neural networks (GNNs) emerge as an intuitive and effective choice for this task. In this study, we evaluate the GNN model for track finding on the data from the GlueX experiment at Jefferson Lab. We use simulation data to train the model and test on both simulation and real GlueX measurements. We demonstrate that GNN-based track finding outperforms the currently used traditional method at GlueX in terms of segment-based efficiency at a fixed purity while providing faster inferences. We show that the GNN model can achieve significant speedup by processing multiple events in batches, which exploits the parallel computation capability of graphical processing units (GPUs). Finally, we compare the GNN implementation on GPU and field-programmable gate array and describe the trade-off.

batched GNN pipeline↗