Search NASA⌕ Search

SEARCH · Search NASA

Results for “GPU computing”

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 127 records · Page 7
All sourcesNASA NTRSDOE OSTI
DOE OSTI2018

Implementation of metal-friendly EAM/FS-type semi-empirical potentials in HOOMD-blue: A GPU-accelerated molecular dynamics software

We present an implementation of EAM and FS interatomic potentials, which are widely used in simulating metallic systems, in HOOMD-blue, a software designed to perform classical molecular dynamics simulations using GPU accelerations. We first discuss the details of our implementation and then report extensive benchmark tests. We demonstrate that single-precision floating point operations efficiently implemented on GPUs can produce sufficient accuracy when compared against double-precision codes, as demonstrated in test simulations of calculations of the glass-transition temperature of Cu 64.5 Zr 35.5 , and pair correlation function of liquid Ni 3 Al. Our code scales well with the size of the simulating system on NVIDIA Tesla M40 and P100 GPUs. Compared with another popular software LAMMPS running on 32 cores of AMD Opteron 6220 processors, the GPU/CPU performance ratio can reach as high as 4.6. In conclusion, the source code can be accessed through the HOOMD-blue web page for free by any interested user.

36 MATERIALS SCIENCE↗
DOE OSTI2019

Fast neural network training on a cluster of GPUs for action recognition with high accuracy

In this work, we propose algorithms and techniques to accelerate training of deep neural networks for action recognition on a cluster of GPUs. The convergence analysis of our algorithm shows it is possible to reduce communication cost and at the same time minimize the number of iterations needed for convergence. We customize the Adam optimizer for our distributed algorithm to improve efficiency. In addition, we employ transfer-learning to further reduce training time while improving validation accuracy. For the UCF101 and HMDB51 datasets, the validation accuracies achieved are 93.1% and 67.9% respectively. With an additional end-to-end trained temporal stream, the validation accuracies achieved for UCF101 and HMDB51 are 93.47% and 81.24% respectively. As far as we know, these are the highest accuracies achieved with the two-stream approach using ResNet that does not involve computationally expensive 3D convolutions or pretraining on much larger datasets.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Low-synch Gram–Schmidt with delayed reorthogonalization for Krylov solvers

The parallel strong-scaling of iterative methods is often determined by the number of global reductions at each iteration. Low-synch Gram-Schmidt algorithms are applied here to the Arnoldi algorithm to reduce the number of global reductions and therefore to improve the parallel strong-scaling of iterative solvers for nonsymmetric matrices such as the GMRES and the Krylov-Schur iterative methods. In the Arnoldi context, the factorization is "left-looking" and processes one column at a time. Among the methods for generating an orthogonal basis for the Arnoldi algorithm, the classical Gram-Schmidt algorithm, with reorthogonalization (CGS2) requires three global reductions per iteration. A new variant of CGS2 that requires only one reduction per iteration is presented and applied to the Arnoldi algorithm. Delayed CGS2 (DCGS2) employs the minimum number of global reductions per iteration (one) for a one-column at-a-time algorithm. The main idea behind the new algorithm is to group global reductions by rearranging the order of operations. DCGS2 must be carefully integrated into an Arnoldi expansion or a GMRES solver. Numerical stability experiments assess robustness for Krylov-Schur eigenvalue computations. Performance experiments on the ORNL Summit supercomputer then establish the superiority of DCGS2 over CGS2.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

Monte Carlo Second- and Third-Order Many-Body Green’s Function Methods with Frequency-Dependent, Nondiagonal Self-Energy

We fully develop the Monte Carlo many-body Green’s function (MC-GF) method with the following enhancements: (1) The truncation order of the perturbation expansion of the Dyson self-energy is raised from the second order (MC-GF2) to the third order (MC-GF3) with the aid of a computerized procedure to enumerate and transform all 84 third-order Goldstone diagrams into Monte Carlo integrable expressions and then into central processing unit (CPU)/graphical processing unit (GPU)-parallel computer codes. (2) An efficient algorithm is proposed that computes all off-diagonal and diagonal elements of the MC-GF2 and MC-GF3 self-energy matrices by common subexpression elimination. (3) The frequency-independent approximation is lifted by introducing a method that computes frequency derivatives of the MC-GF2 and MC-GF3 self-energies up to any arbitrarily high order at nearly no additional computational cost. (4) The imaginary-time integration in the Laplace-transformed expressions of the self-energy is carried out stochastically (instead of using a quadrature in the previous implementations), resulting in a 50- to 200-fold speedup. (5) The efficiency of the redundant-walker convergence acceleration scheme is analyzed numerically, and the guidelines are established to select an optimal number of walkers for maximal efficiency. When such an optimal number is used, the cost per sample is constant of molecular size on either many CPUs or many GPUs. (6) The computational cost to obtain a binding energy within a given statistical uncertainty is observed to increase as (tentatively) O(n 4 ) and O(n 5 ) of molecular size (n is the number of electrons) with and without the diagonal approximation, respectively, as opposed to O(n 5 ) and O(n 6 ) of the corresponding deterministic algorithms. With this method applied to the electron binding energies of C 60 , we show that the third-order corrections to the self-energies are much greater in electron binding energies than in ground-state energies. As a result, they display a sign of oscillatory convergence toward experimental results, not necessarily improving the agreement with increasing perturbation order, justifying MC-GF3 and motivating even higher-order methods.

Algorithms↗
DOE OSTI2025

Coupled Climate Simulations With E3SM‐MMF

Simulations of the recent historical period from 1950 to 2014 are conducted with E3SM‐MMF, which uses an embedded 2D cloud resolving model that runs efficiently on GPUs in place of traditional parameterizations for cloud and turbulence. Analysis of the climate and variability reveal several aspects where E3SM‐MMF produces smaller biases compared to E3SMv2, including better agreement with the observed evolution of global mean surface temperature, although the representation of ENSO is too weak and fast. Three idealized abrupt CO 2 experiments were also conducted to assess climate sensitivity and feedbacks. These yield three estimates of effective climate sensitivity (4.38, 5.21, and 6.06 K), with a corresponding spread in the shortwave cloud feedbacks. These estimates are on the higher end of sensitivity estimates from CMIP ensembles, and the spread indicates substantial state‐dependent feedbacks. These results demonstrate how multiscale modeling framework (MMF) models can be used for climate relevant experiments and projections by leveraging modern GPU enabled computational platforms. The unique qualities of E3SM‐MMF shown in previous literature are largely still present, but various instances of reduced biases suggest that MMF models have utility in improving future projections.

E3SM↗
DOE OSTI2021

Simple and Scalable Streaming: The GRETA Data Pipeline

The Gamma Ray Energy Tracking Array (GRETA) is a state of the art gamma-ray spectrometer being built at Lawrence Berkeley National Laboratory to be first sited at the Facility for Rare Isotope Beams (FRIB) at Michigan State University. A key design requirement for the spectrometer is to perform gamma-ray tracking in near real time. To meet this requirement we have used an inline, streaming approach to signal processing in the GRETA data acquisition system, using a GPU-equipped computing cluster. The data stream will reach 480 thousand events per second at an aggregate data rate of 4 gigabytes per second at full design capacity. We have been able to simplify the architecture of the streaming system greatly by interfacing the FPGA-based detector electronics with the computing cluster using standard network technology. A set of highperformance software components to implement queuing, flow control, event processing and event building have been developed, all in a streaming environment which matches detector performance. Prototypes of all high-performance components have been completed and meet design specifications.

Cromaz, Mario↗
DOE OSTI2022

HyKKT: a hybrid direct-iterative method for solving KKT linear systems

Here, we propose a solution strategy for the large indefinite linear systems arising in interior methods for nonlinear optimization. The method is suitable for implementation on hardware accelerators such as graphical processing units (GPUs). The current gold standard for sparse indefinite systems is the LBLT factorization where L is a lower triangular matrix and B is 1×1 or 2×2 block diagonal. However, this requires pivoting, which substantially increases communication cost and degrades performance on GPUs. Our approach solves a large indefinite system by solving multiple smaller positive definite systems, using an iterative solver on the Schur complement and an inner direct solve (via Cholesky factorization) within each iteration. Cholesky is stable without pivoting, thereby reducing communication and allowing reuse of the symbolic factorization. We demonstrate the practicality of our approach on large optimal power flow problems and show that it can efficiently utilize GPUs and outperform LBL T factorization of the full system.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2017

Möbius domain-wall fermions on gradient-flowed dynamical HISQ ensembles

We report on salient features of a mixed lattice QCD action using valence Möbius domain-wall fermions solved on the dynamical 𝑁 𝑓 = 2 + 1 + 1 highly improved staggered quark sea-quark ensembles generated by the MILC Collaboration. The approximate chiral symmetry properties of the valence fermions are shown to be significantly improved by utilizing the gradient-flow scheme to first smear the highly improved staggered quark configurations. The greater numerical cost of the Möbius domain-wall inversions is mitigated by the highly efficient QUDA library optimized for NVIDIA GPU accelerated compute nodes. We have created an interface to this optimized QUDA solver in Chroma. We provide tuned parameters of the action and performance of QUDA using ensembles with the lattice spacings 𝑎 ≃ {0.15,0.12,0.09} fm and pion masses 𝑚 𝜋 ≃ {310,220,130} MeV. We have additionally generated two new ensembles with 𝑎 ∼ 0.12 fm and 𝑚 𝜋 ∼ {400,350} MeV. Furthermore, with a fixed flow time of 𝑡 𝑔⁡𝑓 =1 in lattice units, the residual chiral symmetry breaking of the valence fermions is kept below 10% of the light quark mass on all ensembles, 𝑚 res ≲ 0.1 × 𝑚 𝑙 , with moderate values of the fifth dimension 𝐿 5 and a domain-wall height 𝑀 5 ≤ 1.3. As a benchmark calculation, we perform a continuum, infinite volume, physical pion and kaon mass extrapolation of 𝐹 𝐾 ±⁡ /𝐹 𝜋 ± and demonstrate our results are independent of flow time and consistent with the FLAG determination of this quantity at the level of less than one standard deviation.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

Dissecting Tensor Cores via Microbenchmarks: Latency, Throughput and Numeric Behaviors

Tensor Cores have been an important unit to accelerate Fused Matrix Multiplication Accumulation (MMA) in all NVIDIA GPUs since Volta Architecture. To program Tensor Cores, users have to use either legacy wmma APIs or current mma APIs. Legacy wmma APIs are more easy-to-use but can only exploit limited features and power of Tensor Cores. Specifically, wmma APIs support fewer operand shapes and can not leverage the new sparse matrix multiplication feature of the newest Ampere Tensor Cores. However, the performance of current programming interface has not been well explored. Furthermore, the computation numeric behaviors of low-precision floating points (TF32, BF16, and FP16) supported by the newest Ampere Tensor Cores are also mysterious. In this paper, we explore the throughput and latency of current programming APIs. Further, we intuitively study the numeric behaviors of Tensor Cores MMA and profile the intermediate operations including multiplication, addition of inner product, and accumulation. All codes used in this work can be found in https://github.com/sunlex0717/DissectingTensorCores.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

A Performance Portable, Fully Implicit Landau Collision Operator with Batched Linear Solvers

Modern accelerators use hierarchical parallel programming models that enable massive multithreading within a processing element (PE), with multiple PEs per device driven by traditional processes. Batching is a technique for exposing PE-level parallelism in algorithms that have traditionally run on MPI processes or multiple threads within a single process. Opportunities for batching arise in, for example, kinetic discretizations of magnetized plasmas where collisions are advanced in velocity space at each spatial point independently. This paper builds on previous work on a high-performance, fully nonlinear, Landau collision operator by batching the linear solver, as well as batching the spatial point problems and adding new support for multiple grids for multiscale, multispecies problems. An anisotropic relaxation verification test that agrees well with previously published results and analytical models is presented. The performance results from NVIDIA A100 and AMD MI250X nodes are presented with hardware utilization analysis for each architecture. Finally, the entire implicit Landau operator time advance is implemented in Kokkos for performance portability, running entirely on the device and is available in the PETSc numerical library.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2017

Acceleration of the IMplicit–EXplicit nonhydrostatic unified model of the atmosphere on manycore processors

We present the acceleration of an IMplicit-EXplicit (IMEX) nonhydrostatic atmospheric model on manycore processors such as graphic processing units (GPUs) and Intel's Many Integrated Core (MIC) architecture. IMEX time integration methods sidestep the constraint imposed by the Courant-Friedrichs-Lewy condition on explicit methods through corrective implicit solves within each time step. In this work, we implement and evaluate the performance of IMEX on manycore processors relative to explicit methods. Using 3D-IMEX at Courant number C = 15, we obtained a speedup of about 4x relative to an explicit time stepping method run with the maximum allowable C = 1. Moreover, the unconditional stability of IMEX with respect to the fast waves means the speedup can increase significantly with the Courant number as long as the accuracy of the resulting solution is acceptable. We show a speedup of 100x at C = 150 using 1D-IMEX to demonstrate this point. Several improvements on the IMEX procedure were necessary in order to outperform our results with explicit methods: (a) reducing the number of degrees of freedom of the IMEX formulation by forming the Schur complement, (b) formulating a horizontally explicit vertically implicit 1D-IMEX scheme that has a lower workload and better scalability than 3D-IMEX, (c) using high-order polynomial preconditioners to reduce the condition number of the resulting system, and (d) using a direct solver for the 1D-IMEX method by performing and storing LU factorizations once to obtain a constant cost for any Courant number. Without all of these improvements, explicit time integration methods turned out to be difficult to beat. We discuss in detail the IMEX infrastructure required for formulating and implementing efficient methods on manycore processors. Several parametric studies are conducted to demonstrate the gain from each of the abovementioned improvements. Lastly, we validate our results with standard benchmark problems in numerical weather prediction and evaluate the performance and scalability of the IMEX method using up to 4192 GPUs and 16 Knights Landing processors.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

EdgeAI: Machine learning via direct attached accelerator for streaming data processing at high shot rate x-ray free-electron lasers

We present a case for low batch-size inference with the potential for adaptive training of a lean encoder model. We do so in the context of a paradigmatic example of machine learning as applied in data acquisition at high data velocity scientific user facilities such as the Linac Coherent Light Source-II x-ray Free-Electron Laser. We discuss how a low-latency inference model operating at the data acquisition edge can capitalize on the naturally stochastic nature of such sources. We simulate the method of attosecond angular streaking to produce representative results whereby simulated input data reproduce high-resolution ground truth probability distributions. By minimizing the mean-squared error between the decoded output of the latent representation and the ground truth distributions, we ensure that the encoding layers and resulting latent representation maintains full fidelity for any downstream task, be it classification or regression. We present throughput results for data-parallel inference of various batch sizes, some with throughput exceeding 100 k images per second. We also show in situ training below 10 s per epoch for the full encoder–decoder model as would be relevant for streaming and adaptive real-time data production at our nation’s scientific light sources.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

GIGA-Lens: Fast Bayesian Inference for Strong Gravitational Lens Modeling

We present GIGA-Lens: a gradient-informed, GPU-accelerated Bayesian framework for modeling strong gravitational lensing systems, implemented in TensorFlow and JAX. The three components, optimization using multistart gradient descent, posterior covariance estimation with variational inference, and sampling via Hamiltonian Monte Carlo, all take advantage of gradient information through automatic differentiation and massive parallelization on graphics processing units (GPUs). We test our pipeline on a large set of simulated systems and demonstrate in detail its high level of performance. The average time to model a single system on four Nvidia A100 GPUs is 105 s. The robustness, speed, and scalability offered by this framework make it possible to model the large number of strong lenses found in current surveys and present a very promising prospect for the modeling of ${ \mathcal O }({10}^{5})$ lensing systems expected to be discovered in the era of the Vera C. Rubin Observatory, Euclid, and the Nancy Grace Roman Space Telescope.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Comparing Early Evolution of Flames in X-Ray Bursts in Two and Three Dimensions

We explore the early evolution of flame ignition and spreading on the surface of a neutron star in three dimensions, in the context of X-ray bursts. We look at the nucleosynthesis and morphology of the burning front and compare to two-dimensional axisymmetric simulations to gauge how important a full three-dimensional treatment of the flame is for the early dynamics. Finally, we discuss the progress toward full-star resolved flame simulations.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2024

Strong Coupling of Hydrodynamics and Reactions in Nuclear Statistical Equilibrium for Modeling Convection in Massive Stars

We build on the simplified spectral deferred corrections (SDC) coupling of hydrodynamics and reactions to handle the case of nuclear statistical equilibrium (NSE) and electron/positron captures/decays in the cores of massive stars. Our approach blends a traditional reaction network on the grid with a tabulated NSE state from a very large, ${\mathcal O }(100)$ nuclei network. We demonstrate how to achieve second-order accuracy in the simplified-SDC framework when coupling NSE to hydrodynamics, with the ability to evolve the star on the hydrodynamics time step. We discuss the application of this method to convection in massive stars leading up to core collapse. We also show how to initialize the initial convective state from a 1D model in a self-consistent fashion. All of these developments are done in the publicly available Castro simulation code and the entire simulation methodology is fully GPU-accelerated.

Explosive nucleosynthesis↗
DOE OSTI2025

Numerical Treatment of Shock-induced Nuclear Burning in Double Detonation Type Ia Supernovae

We present a benchmark problem to assess the treatment of shock-induced nuclear burning in the context of double detonation Type Ia supernovae. In a stratified white dwarf model, we implement a shock-detection criterion that suppresses burning in zones characterized by compression and significant pressure gradients, controlled by a tunable parameter, f shock . One-dimensional simulations, using the open-source Castro suite, were conducted across three treatments—burning fully enabled, and burning suppressed with f shock = 2/3 and f shock = 1—across three spatial resolutions (5.0, 2.5, and 0.3125 km). At the finest resolution, the burning-enabled and f shock = 1 models converge, while the f shock = 2/3 front continues to show slight offset behavior. Since most simulations are carried out at much lower resolutions, our tests support the idea that burning in shocks should always be disabled in practice. We also observe that the behavior of lower-resolution simulations remains extremely sensitive to the choice of f shock .

GPU computing↗
NASA NTRS2012

Holodeck: Telepresence Dome Visualization System Simulations

This paper explores the simulation and consideration of different image-projection strategies for the Holodeck, a dome that will be used for highly immersive telepresence operations in future endeavors of the National Aeronautics and Space Administration (NASA). Its visualization system will include a full 360 degree projection onto the dome's interior walls in order to display video streams from both simulations and recorded video. Because humans innately trust their vision to precisely report their surroundings, the Holodeck's visualization system is crucial to its realism. This system will be rigged with an integrated hardware and software infrastructure-namely, a system of projectors that will relay with a Graphics Processing Unit (GPU) and computer to both project images onto the dome and correct warping in those projections in real-time. Using both Computer-Aided Design (CAD) and ray-tracing software, virtual models of various dome/projector geometries were created and simulated via tracking and analysis of virtual light sources, leading to the selection of two possible configurations for installation. Research into image warping and the generation of dome-ready video content was also conducted, including generation of fisheye images, distortion correction, and the generation of a reliable content-generation pipeline.

Hite, Nicolas↗
NASA NTRS2016

AIRNOISE: A Tool for Preliminary Noise-Abatement Terminal Approach Route Design

Noise from aircraft in the airport vicinity is one of the leading aviation-induced environmental issues. The FAA developed the Integrated Noise Model (INM) and its replacement Aviation Environmental Design Tool (AEDT) software to assess noise impact resulting from all aviation activities. However, a software tool is needed that is simple to use for terminal route modification, quick and reasonably accurate for preliminary noise impact evaluation and flexible to be used for iterative design of optimal noise-abatement terminal routes. In this paper, we extend our previous work on developing a noise-abatement terminal approach route design tool, named AIRNOISE, to satisfy this criterion. First, software efficiency has been significantly increased by over tenfold using the C programming language instead of MATLAB. Moreover, a state-of-the-art high performance GPU-accelerated computing module is implemented that was tested to be hundreds time faster than the C implementation. Secondly, a Graphical User Interface (GUI) was developed allowing users to import current terminal approach routes and modify the routes interactively to design new terminal approach routes. The corresponding noise impacts are then calculated and displayed in the GUI in seconds. Finally, AIRNOISE was applied to Baltimore-Washington International Airport terminal approach route to demonstrate its usage.

aircraft noise↗