Search NASA⌕ Search

SEARCH · Search NASA

Results for “NVIDIA GPU”

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 271 records · Page 15
All sourcesNASA NTRSDOE OSTI
NASA NTRSDate not supplied

Implicit Thermochemical Nonequilibrium Flow Simulations on Unstructured Grids using GPUs

Thermochemical nonequilibrium flow simulation capabilities have been previously implemented, verified, and validated for central processing unit (CPU) systems in NASA’s unstructured-grid computational fluid dynamics solver FUN3D. Many exascale-class high-performance computing systems will rely on graphics processing unit (GPU) architectures for high throughput and energy efficiency; thus, CPU-based scientific computing software unable to effectively utilize these systems must be updated. In this work, we present a CUDA C++ implementation of FUN3D’s thermochemical nonequilibrium flow simulation capabilities targeting NVIDIA Tesla GPUs. An overview of the porting and optimization strategy is described and performance comparisons with other recent architectures are presented. Scaling to thousands of GPUs is demonstrated, yielding computational performance equivalent to that of several million CPU cores. The implementation enables efficient, high-fidelity, scale-resolving simulations of thermochemical nonequilibrium flows for many applications including atmospheric entry, hypersonics, and combustion.

GPU↗
DOE OSTI2026

How efficiently can AI recognize Wireless Devices?

This poster presents a hardware benchmarking methodology for a 3-layer CNN waveform classifier deployed using ONNX Runtime on an NVIDIA Jetson AGX Orin. The dataset consist of 9 signal types, -30 to +30 dB SNR with 5dB increments. Benchmarking on the Jetson AGX Orin gave an accuracy of 91.9% and GPU throughput of 107,120 predictions/sec (23× faster than CPU). The Jetson GPU reached approximately 27M samples/sec with stable performance but fell below the 40 MHz rate needed for real-time radio feeds. Sustained testing of 5 minutes confirmed stable performance with no memory leaks, establishing a reproducible benchmarking baseline for future edge-deployment optimization.

99 - GENERAL AND MISCELLANEOUS↗
DOE OSTI2022

m-CUBES An efficient and portable implementation of multi-dimensional integration for gpus

The task of multi-dimensional numerical integration is frequently encountered in physics and other scientific fields, e.g., in modeling the effects of systematic uncertainties in physical systems and in Bayesian parameter estimation. Multi-dimensional integration is often time-prohibitive on CPUs. Efficient implementation on many-core architectures is challenging as the workload across the integration space cannot be predicted a priori. We propose m-Cubes, a novel implementation of the well-known Vegas algorithm for execution on GPUs. Vegas transforms integration variables followed by calculation of a Monte Carlo integral estimate using adaptive partitioning of the resulting space. m-Cubes improves performance on GPUs by maintaining relatively uniform workload across the processors. As a result, our optimized Cuda implementation for Nvidia GPUs outperforms parallelization approaches proposed in past literature. We further demonstrate the efficiency of m-Cubes by evaluating a six-dimensional integral from a cosmology application, achieving significant speedup and greater precision than the CUBA library's CPU implementation of VEGAS. We also evaluate m-Cubes on a standard integrand test suite. m-Cubes outperforms the serial implementations of the Cuba and GSL libraries by orders of magnitude speedup while maintaining comparable accuracy. Our approach yields a speedup of at least 10 when compared against publicly available Monte Carlo based GPU implementations. In summary, m-Cubes can solve integrals that are prohibitively expensive using standard libraries and custom implementations. A modern C++ interface header-only implementation makes m-Cubes portable, allowing its utilization in complicated pipelines with easy to define stateful integrals. Compatibility with non-Nvidia GPUs is achieved with our initial implementation of m-Cubes using the Kokkos framework.

Sakiotis, Ioannis↗
DOE OSTI2021

Performance Profile of Transformer Fine-Tuning in Multi-GPU Cloud Environments

The study presented here focuses on performance characteristics and trade-offs associated with running machine-learning tasks in multi-GPU environments on both on-site cloud computing resources and commercial cloud services (Azure). Specifically, this study examines these tradeoffs by examining the performance of training and fine-tuning of transformer-based deep-learning (DL) networks on clinical notes and data, a task of critical importance in the medical domain. To this end, we perform DL-related experiments on the widely deployed NVIDIA V100 GPUs and on the newer A100 GPUs connected via NVLink or PCIe. This study analyzes the execution time of major operations to train DL models and investigate popular options to optimize each of them. We examine and present the findings on the impacts that various operations (e.g. data loading into GPUs, training, fine-tuning), optimizations, and system configurations (single vs. multi-GPU, NVLink vs. PCIe) have on the overall training performance.

Begoli, Edmon↗
DOE OSTI2024

Analyzing inference workloads for spatiotemporal modeling

Ensuring power grid resiliency, forecasting climate conditions, and optimization of transportation infrastructure are some of the many application areas where data is collected in both space and time. Spatiotemporal modeling is about modeling those patterns for forecasting future trends and carrying out critical decision-making by leveraging machine learning/deep learning. Once trained offline, field deployment of trained models for near real-time inference could be challenging because performance can vary significantly depending on the environment, available compute resources and tolerance to ambiguity in results. Users deploying spatiotemporal models for solving complex problems can benefit from analytical studies considering a plethora of system adaptations to understand the associated performance-quality trade-offs. To facilitate the co-design of next-generation hardware architectures for field deployment of trained models, it is critical to characterize the workloads of these deep learning (DL) applications during inference and assess their computational patterns at different levels of the execution stack. In this paper, we develop several variants of deep learning applications that use spatiotemporal data from dynamical systems. We study the associated computational patterns for inference workloads at different levels, considering relevant models (Long short-term Memory, Convolutional Neural Network and Spatio-Temporal Graph Convolution Network), DL frameworks (Tensorflow and PyTorch), precision (FP16, FP32, AMP, INT16 and INT8), inference runtime (ONNX and AI Template), post-training quantization (TensorRT) and platforms (Nvidia DGX A100 and Sambanova SN10 RDU). Overall, our findings indicate that although there is potential in mixed-precision models and post-training quantization for spatiotemporal modeling, extracting efficiency from contemporary GPU systems might be challenging. Instead, co-designing custom accelerators by leveraging optimized High Level Synthesis frameworks (such as SODA High-Level Synthesizer for customized FPGA/ASIC targets) can make workload-specific adjustments to enhance the efficiency.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Multiphysics analysis system for heat pipe cooled micro-reactors employing PRAGMA-OpenFOAM-ANLHTP

A multiphysics analysis system for neutronics/thermo-mechanical/heat-pipe analysis of heat pipe cooled micro-reactors was developed using the PRAGMA code as the neutronics engine. PRAGMA, which used to be a GPU-based continuous-energy MC code for power reactor applications, now has an extended geometry package to handle geometries with unstructured meshes generated by the ANSYS Design-Modeler and Meshing. The NVIDIA ray tracing engine OptiX was exploited for efficient neutron transport on unstructured geometry. On the multiphysics side, the open-source CFD tool OpenFOAM and one-dimensional heat pipe analysis code ANLHTP were adopted. The manager-worker system based on the MPI dynamic process management (DPM) model enables efficient coupling of codes employing different parallelization schemes. With all the features, the multiphysics analysis of the one-sixth symmetrical MegaPower 2D core was performed and it demonstrated the benefit of the tight integration of the three-way coupling system and one-to-one geometry coupling strategy. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
NASA NTRS2016

Utilizing GPUs to Accelerate Turbomachinery CFD Codes

GPU computing has established itself as a way to accelerate parallel codes in the high performance computing world. This work focuses on speeding up APNASA, a legacy CFD code used at NASA Glenn Research Center, while also drawing conclusions about the nature of GPU computing and the requirements to make GPGPU worthwhile on legacy codes. Rewriting and restructuring of the source code was avoided to limit the introduction of new bugs. The code was profiled and investigated for parallelization potential, then OpenACC directives were used to indicate parallel parts of the code. The use of OpenACC directives was not able to reduce the runtime of APNASA on either the NVIDIA Tesla discrete graphics card, or the AMD accelerated processing unit. Additionally, it was found that in order to justify the use of GPGPU, the amount of parallel work being done within a kernel would have to greatly exceed the work being done by any one portion of the APNASA code. It was determined that in order for an application like APNASA to be accelerated on the GPU, it should not be modular in nature, and the parallel portions of the code must contain a large portion of the code's computation time.

computer programming↗
DOE OSTI2018

Scaling Deep Learning workloads: NVIDIA DGX-1/Pascal and Intel Knights Landing

Deep Learning (DL) algorithms have become ubiquitous in data analytics. As a result, major computing vendors—including NVIDIA, Intel, AMD, and IBM—have architectural road maps influenced by DL workloads. Furthermore, several vendors have recently advertised new computing products as accelerating large DL workloads. Unfortunately, it is difficult for data scientists to quantify the potential of these different products. Here, this article provides a performance and power analysis of important DL workloads on two major parallel architectures: NVIDIA DGX-1 (eight Pascal P100 GPUs interconnected with NVLink) and Intel Knights Landing (KNL) CPUs interconnected with Intel Omni-Path or Cray Aries. Our evaluation consists of a cross section of convolutional neural net workloads: CifarNet, AlexNet, GoogLeNet, and ResNet50 topologies using the Cifar10 and ImageNet datasets. The workloads are vendor-optimized for each architecture. We use sequentially equivalent implementations to maintain iso-accuracy between parallel and sequential DL models. Our analysis indicates that although GPUs provide the highest overall performance, the gap can close for some convolutional networks; and the KNL can be competitive in performance/watt. We find that NVLink facilitates scaling efficiency on GPUs. However, its importance is heavily dependent on neural network architecture. Furthermore, for weak-scaling—sometimes encouraged by restricted GPU memory—NVLink is less important.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Multinode Multi-GPU Two-Electron Integrals: Code Generation Using the Regent Language

The computation of two-electron repulsion integrals (ERIs) is often the most expensive step of integral-direct self-consistent field methods. Formally it scales as O(N 4 ), where N is the number of Gaussian basis functions used to represent the molecular wave function. In practice, this scaling can be reduced to O(N 2 ) or less by neglecting small integrals with screening methods. The contributions of the ERIs to the Fock matrix are of Coulomb (J) and exchange (K) type and require separate algorithms to compute matrix elements efficiently. We previously implemented highly efficient GPU-accelerated J-matrix and K-matrix algorithms in the electronic structure code TeraChem. Although these implementations supported the use of multiple GPUs on a node, they did not support the use of multiple nodes. This presents a key bottleneck to cutting-edge ab initio simulations of large systems, e.g., excited state dynamics of photoactive proteins. We present our implementation of multinode multi-GPU J- and K-matrix algorithms in TeraChem using the Regent programming language. Regent directly supports distributed computation in a task-based model and can generate code for a variety of architectures, including NVIDIA GPUs. We demonstrate multinode scaling up to 45 GPUs (3 nodes) and benchmark against hand-coded TeraChem integral code. Finally, we also outline our metaprogrammed Regent implementation, which enables flexible code generation for integrals of different angular momenta.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Reproduced Computational Results Report for “Ginkgo: A Modern Linear Operator Algebra Framework for High Performance Computing”

The article titled “Ginkgo: A Modern Linear Operator Algebra Framework for High Performance Computing” by Anzt et al. presents a modern, linear operator centric, C++ library for sparse linear algebra. Experimental results in the article demonstrate that Ginkgo is a flexible and user-friendly framework capable of achieving high-performance on state-of-the-art GPU architectures. In this report, the Ginkgo library is installed and a subset of the experimental results are reproduced. Specifically, the experiment that shows the achieved memory bandwidth of the Ginkgo Krylov linear solvers on NVIDIA A100 and AMD MI100 GPUs is redone and the results are compared to what presented in the published article. Upon completion of the comparison, the published results are deemed reproducible.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Sum Reduction with OpenMP Offload on NVIDIA Grace-Hopper System

We evaluate the performance of the baseline and optimized reductions in OpenMP on an NVIDIA Grace-Hopper system. We explore the impacts of the number of teams, the number of elements to sum per loop iteration, and simultaneous execution on the central-processing unit (CPU) and the GPU in the unified memory (UM) mode upon the reduction performance. The experimental results show that the optimized reductions are 6.120X to 20.906X faster than the baselines on the GPU, and their efficiency ranges from 89% to 95% of the theoretical GPU memory bandwidth. Depending on where an input array is allocated in the program when co-running the reduction on the CPU and GPU in the UM mode, the average speedup over the GPU-only execution is approximately 2.484 or 1.067, and the speedup of the optimized reductions over the baseline reductions ranges from 0.996 to 10.654 or from 0.998 to 6.729.

Jin, Zheming↗
DOE OSTI2025

Enabling Multireference Calculations on Multimetallic Systems with Graphic Processing Units

Modeling multimetallic systems efficiently enables faster prediction of desirable chemical properties and the design of new materials. This work describes an initial implementation for performing multireference wave function method localized active-space self-consistent field (LASSCF) calculations through the use of multiple graphics processing units (GPUs) to accelerate time-to-solution. Density fitting is leveraged to reduce memory requirements, and we demonstrate the ability to fully utilize multi-GPU compute nodes. Performance improvements of 5–10x in total application runtime were observed in LASSCF calculations for multimetallic catalyst systems up to 1200 AOs and an active space of (22e,40o) using up to four NVIDIA A100 GPUs. Furthermore, written with performance portability in mind, a comparable performance is also observed in early runs on the Aurora exascale system using Intel Max Series GPUs.

Algorithms↗
DOE OSTI2024

FTTN: Feature-Targeted Testing for Numerical Properties of NVIDIA & AMD Matrix Accelerators

While NVIDIA has been the dominant provider of GPUs for HPC and ML, now AMD has several offerings of GPUs. This encourages programmers to try out AMD GPUs for new codes and also port existing codes over. Unfortunately, without understanding the floating-point differences between these GPU types, software development or porting can introduce bugs—and currently such an understanding is lacking. The magnitude of this open question becomes clear if one imagines the the number of floating-point precision choices (FP16, FP32, etc.), floating-point formats (standard floats, brain-float, etc.), and execution units available (elementary units, matrix/tensor cores, etc.) Questions such as rounding modes and subnormal support are also important. Most of these answers are unknown today or are hard to access. We provide the first testing-guided approach that answers a significant number of these questions. We also devise tests to reveal internal information (e.g., extra bits kept) to make sure that our findings are reliable. Many of our tests employ systematically generated random-programs, others apply fast-math flags and some involve fused multiplyadd. Especially for tensor/matrix cores, the tests have nontrivial logic that we present Our testing approach is reusable for the plethora of GPUs yet to be introduced. Our findings include up to 7 ulps of difference between NVIDIA and AMD for sin and cos at FP32 precision and 3 ulp at FP64. In our study of matrix cores (NVIDIA) and tensor cores (AMD), we have extensively characterized rounding modes (truncation versus round-to-nearest), the number of extra internal bits kept (whether 3 bits are kept or not), subnormal support for inputs and outputs across four different floating-point formats and across NVIDIA A100 and AMD MI250X GPUs. We believe that this wealth of data becoming available for the first time may help avoid significant porting bugs when migrating code across these platforms.

Li, Xinyi↗
DOE OSTI2023

Studying performance portability of LAMMPS across diverse GPU-based platforms

The molecular dynamics simulation software, LAMMPS, utilizes the Kokkos acceleration library to port computation to a diverse set of architectures including those based on GPU accelerators. In addition to Kokkos, LAMMPS contains a vast code base that leverages the CUDA application programming interface using library functions such as cuFFT, CUDA's fast-fourier transform (FFT) library, and, more recently, also support for AMD's Heterogeneous Interface for Portability (HIP) that is rapidly growing. While preparing LAMMPS tests for the AMD GPU-based test system precursors to Frontier, we investigated several strategies for accelerating LAMMPS on AMD GPUs, using the AMD Instinct MI100 and MI250X. In this work, we integrated the HIP FFT library, hipFFT, into the particle-particle particle-mesh (PPPM) long-range solver, which allowed the porting of PPPM calculations to the GPUs. Kokkos behavior on the MI100 and MI250X was also investigated through the package kokkos command of LAMMPS, targeting communication, memory usage, and particle grid decomposition. The Tersoff, Reax, Lennard-Jones (LJ), EAM, Granular, and PPPM potentials were investigated in this effort, and results from these experiments are provided. In conclusion, the selected potentials were run on Spock (AMD Instinct MI100), Crusher (AMD Instinct MI250X), AFW HPC11 (NVIDIA A100) and Summit (NVIDIA V100), for comparison. Operational roofline models were constructed and analyzed for the Tersoff, Reax, and Lennard–Jones potentials on Crusher and Summit.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Nek5000/RS Performance on Advanced GPU Architectures

We demonstrate NekRS performance results on various advanced GPU architectures. NekRS is a GPU-accelerated version of Nek5000 that is targeting high performance on forthcoming exascale platforms. It is being developed in DOE’s Center of Efficient Exascale Discretizations (CEED), which is one of the co-design centers under the Exascale Computing Project (ECP). In this report, we consider Frontier, Crusher, Spock, Polaris, Perlmutter, ThetaGPU and Summit. Simulations are performed using ExaSMR’s 17x17 rod-bundle geometries with different problem sizes. The report focuses on strong scaling performance and analysis. Many of the results shown in this report are the outcome from participation in the ALCF GPU Hackathon on Polaris, which was held on 7/19/22, and 7/26–7/28/22. Members of the Nek5000/RS team included Misun Min, Yu-Hsiang Lan, Paul Fischer and Thilina Rathnayake. Mentors were Kris Rowe (ALCF) and Peng Wang (NVIDIA). Also presented in this report are results on Frontier, which were obtained in collaboration with John Holmen at OLCF.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

GPU acceleration of Swendsen–Wang dynamics

When simulating a lattice system near its critical temperature, local algorithms for modeling the system’s evolution can introduce very large autocorrelation times into sampled data. Here, this critical slowing down places restrictions on the analysis that can be completed in a timely manner of the behavior of systems around the critical point. Because it is often desirable to study such systems around this point, a new algorithm must be introduced. Therefore, we turn to cluster algorithms, such as the Swendsen–Wang algorithm and the Wolff clustering algorithm. They incorporate global updates which generate new lattice configurations with little correlation to previous states, even near the critical point. We look to accelerate the rate at which these algorithm are capable of running by implementing and benchmarking a parallel implementation of each algorithm designed to run on GPUs under NVIDIA’s CUDA framework. A 17 and 90 fold increase in the computational rate was, respectively, experienced when measured against the equivalent algorithm implemented in serial code.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2023

The Kokkos OpenMPTarget Backend: Implementation and Lessons Learned

As the supercomputing landscape diversifies, solutions such as Kokkos to write vendor agnostic applications and libraries have risen in popularity. Kokkos provides a programming model designed for performance portability, which allows developers to write a single source implementation that can run efficiently on various architectures. At its heart, Kokkos maps parallel algorithms to architecture and vendor specific backends written in lower level programming models such as CUDA and HIP. Another approach to writing vendor agnostic parallel code is using OpenMP’s directives based approach, which lets developers annotate code to express parallelism. It is implemented at the compiler level and is supported by all major high performance computing vendors, as well as the primary Open Source toolchains GNU and LLVM. Since its inception, Kokkos has used OpenMP to parallelize on CPU architectures. In this paper, we explore leveraging OpenMP for a GPU backend and discuss the challenges we encountered when mapping the Kokkos APIs and semantics to OpenMP target constructs. As an exemplar workload we chose a simple conjugate gradient solver for sparse matrices. We find that performance on NVIDIA and AMD GPUs varies widely based on details of the implementation strategy and the chosen compiler. Furthermore, the performance of the OpenMP implementations decreases with increasing complexity of the investigated algorithms.

Gayatri, Rahulkumar↗