An Evaluative Comparison of Performance Portability across GPU Programming Models
Explore the source record for details and available documents.
Computer science↗
SEARCH · Search NASA
Results for “GPU performance modeling”
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.
Taking GPU Programming Models to Task for Performance Portability
Explore the source record for details and available documents.
Computer science↗
An automated and portable method for selecting an optimal GPU frequency
Power consumption poses a significant challenge in current and emerging graphics processing unit (GPU) enabled high-performance computing systems. In modern GPUs, dynamic voltage frequency scaling (DVFS) appears to be a reliable control to regulate power consumption and performance. However, the DVFS design space is large - hence, brute-force approaches are infeasible to select the optimal frequency. Furthermore, no single frequency can be universally optimal for applications with varying computational intensities. Thus, the application's complexity and the availability of a wide range of frequency settings are a challenge in selecting the optimal frequency configuration for a given GPU workload. To that end, this paper proposes a systematic approach that consists of three steps. The feature characterization study identifies the fine-grain GPU utilization metrics that influence the power consumption and execution time of a given workload. To understand the performance, power, and energy consumption behaviors of a workload across GPU's DVFS design space, we derived analytical power and performance models using the identified fine-grain features. Here, it is shown that the same set of GPU utilization metrics can estimate both the power consumption and execution time while being agnostic of changes to frequency and input sizes. Applying a power control with the single objective of reducing power may cause performance degradation, leading to more energy consumption. A multi-objective approach is proposed to select the optimal GPU DVFS configuration for a workload that reduces power consumption with negligible degradation in performance. The evaluation was conducted using SPEC ACCEL benchmarks and three real applications - NAMD LAMMPS, and LSTM on NVIDIA GV100, GA100, and AMD MI210 GPUs. On average, real applications showed 29.6% energy savings with a performance loss of 5.2% on GA100 and 22.6% energy savings with a performance loss of 4.7% on GV100. Moreover, the proposed models are portable to real applications, GPU architectures, and vendors, and require metric collection at only the default frequency rather than all supported DVFS configurations. Additionally, we conducted a comparison between our models and the GPU assembly instructions (PTX)-based static models. The results revealed a significant reduction in the average error rates, with a decrease from 19.7% to 3.1% for power models and from 29.4% to 5.2% for performance models.
97 MATHEMATICS AND COMPUTING↗
Did You Win the GPU Cloud Lottery? Benchmarking from TFLOPS to Tokens/$
Cloud GPUs are commonly assumed to deliver consistent performance for a given GPU model. This assumption does not always hold: cloud providers employ diverse system configurations and virtualization mechanisms, and GPUs themselves exhibit non-negligible manufacturing variability (the silicon lottery). In this work, we present a large-scale measurement study of GPU performance variability across 11 cloud providers, covering over 3,500 physical GPUs and 6,800 benchmark runs. Our hierarchical analysis shows that while execution-level variation stays below 9%, performance varies by up to 38% across devices and providers for the same GPU model. Regression analysis indicates that driver- and OS-related software factors contribute less than 1% of the variance; instead, silicon lottery effects dominate observed performance variation, and cloud providers further amplify them through persistent, systematic second-order effects.
Slynko, Platon [Silicon Data, New York, USA] (ORCI↗
GPU-friendly surface model for Monte-Carlo detector simulations
The demands for Monte-Carlo simulation are drastically increasing with the Large Hadron Collider’s high-luminosity upgrade, and are expected to exceed the currently available compute resources. At the same time, modern high-performance computing has adopted powerful hardware accelerators, particularly GPUs. The AdePT and Celeritas projects aim to address the demanding computational needs by leveraging these heterogeneous computing architectures. While both have successfully ported realistic detector simulations to GPUs using the VecGeom library, the complexity of geometry modeling emerged as a bottleneck. Thread divergence and high register usage were degrading the GPU performance. Therefore, a new, GPU-friendly surface-based model has been introduced in the VecGeom library that decomposes the divergent code of the 3D primitive solids into simpler and more balanced surface algorithms. In this work, we present the latest developments, focusing on the additions required to efficiently model complex setups like the CMS Phase-2 geometry. This includes memory reduction techniques, and adding accelerating structures for faster traversal.
Diederichs, Severin [CERN]↗
Harnessing distributed GPU computing for generalizable graph convolutional networks in power grid reliability assessments
Although machine learning (ML) has emerged as a powerful tool for rapidly assessing grid contingencies, prior studies have largely considered a static grid topology in their analyses. This limits their application, since they need to be re-trained for every new topology. Here, this paper explores the development of generalizable graph convolutional network (GCN) models by pre-training them across a range of grid topologies and contingency types. We found that a GCN model with auto-regressive moving average (ARMA) layers with a line graph representation of the grid offered the best predictive performance in predicting voltage magnitudes (VM) and voltage angles (VA). We introduced the concept of phantom nodes to consider disparate grid topologies with a varying number of nodes and lines. For pre-training the GCN ARMA model across a variety of topologies, distributed graphics processing unit (GPU) computing afforded us significant training scalability. The predictive performance of this model on grid topologies that were part of the training data is substantially better than the direct current (DC) approximation. Although direct application of the pre-trained model to topologies that are not part of the grid is not particularly satisfactory, fine-tuning with small amounts of data from a specific topology of interest significantly improves predictive performance. In general, this paper highlights the feasibility of training large-scale GNN models to assess the reliability of power grids by considering a wide variety of grid topologies and contingency types. With the advent of foundational models in ML and the exponential increase in GPU computing clusters, generalizable ML models will significantly enhance how utilities manage power systems and make decisions in real-time or near-real-time.
24 - POWER TRANSMISSION AND DISTRIBUTION↗
Implementation and Validation of a Generalized Actuator Disk Parameterization for Wind Turbine Simulations Within the FastEddy Model
ABSTRACT Fast and accurate large‐eddy simulation (LES) of the atmospheric boundary layer plays a crucial role in advancing wind energy research. Long‐duration wind farm studies at turbine‐resolving scales have become increasingly important to understand the intricate interactions between large wind farms and the atmospheric boundary layer. However, the prohibitive computational cost of these turbulence‐ and turbine‐resolving simulations has precluded such modeling to be exercised on a regular basis. To that end, we implement and validate the generalized actuator disk (GAD) model in the computationally efficient, graphics processing unit (GPU)–resident, LES model FastEddy. We perform single‐turbine simulations under three atmospheric stabilities (neutral, unstable, and stable) and compare them against observations from the Scaled Wind Farm Technology (SWiFT) facility and other LES codes from the recent Wakebench turbine wake model benchmark. Our idealized LES results agree well with observed wake velocity deficit and downstream recovery across stability regimes. Turbine response in terms of rotational speed, generated power, torque, and thrust coefficient are well predicted across stability regimes and are consistent with the LES results from the benchmark. The FastEddy simulations are found to be at least two orders of magnitude more efficient than the traditional CPU‐based LES models, opening the door for realistic LES simulations of full wind plants as a viable standard practice.
17 WIND ENERGY↗
High Performance, High Fidelity: A GPU‐Accelerated Doubly‐Periodic Configuration of the Simple Cloud‐Resolving E3SM Atmosphere Model Version 1 (DP‐SCREAMv1)
The development of the Simplified Cloud Resolving Energy Exascale Earth System Atmosphere Model (SCREAMv1) enables global storm-resolving simulations on modern GPU-based supercomputers. However, the high computational cost of SCREAMv1 limits its routine use for process-level studies, creating a need for efficient proxy configurations. This study addresses this gap by introducing DP-SCREAMv1, a doubly periodic cloud-resolving model designed to be fully consistent with SCREAMv1 while enabling high-resolution, long-duration simulations at significantly reduced computational expense by simulating a limited doubly periodic domain rather than the entire globe. Built on a C++/Kokkos architecture, DP-SCREAMv1 achieves exceptional performance scalability on GPU systems and includes a rich library of cases for validation and scientific exploration. In this work, we demonstrate short wall-clock times at SCREAMv1's default resolution and show that DP-SCREAMv1 supports routine execution of large-domain, high-resolution experiments that were previously challenging in practice. Furthermore, we show that DP-SCREAMv1 enables routine execution of “Giga-LES” style simulations and facilitates large-domain, high-resolution simulations that were recently considered burdensome to perform. These results document an efficient, fully consistent process-level configuration for SCREAMv1 (DP-SCREAMv1) and illustrate its use for long-duration and large-domain experiments at cloud-resolving to eddy-permitting resolution.
Environmental sciences↗
High-Performance Modeling and Simulation of Anchoring in Granular Media for NEO Applications
NASA is interested in designing a spacecraft capable of visiting a near-Earth object (NEO), performing experiments, and then returning safely. Certain periods of this mission would require the spacecraft to remain stationary relative to the NEO, in an environment characterized by very low gravity levels; such situations require an anchoring mechanism that is compact, easy to deploy, and upon mission completion, easy to remove. The design philosophy used in this task relies on the simulation capability of a high-performance multibody dynamics physics engine. On Earth, it is difficult to create low-gravity conditions, and testing in low-gravity environments, whether artificial or in space, can be costly and very difficult to achieve. Through simulation, the effect of gravity can be controlled with great accuracy, making it ideally suited to analyze the problem at hand. Using Chrono::Engine, a simulation pack age capable of utilizing massively parallel Graphic Processing Unit (GPU) hardware, several validation experiments were performed. Modeling of the regolith interaction has been carried out, after which the anchor penetration tests were performed and analyzed. The regolith was modeled by a granular medium composed of very large numbers of convex three-dimensional rigid bodies, subject to microgravity levels and interacting with each other with contact, friction, and cohesional forces. The multibody dynamics simulation approach used for simulating anchors penetrating a soil uses a differential variational inequality (DVI) methodology to solve the contact problem posed as a linear complementarity method (LCP). Implemented within a GPU processing environment, collision detection is greatly accelerated compared to traditional CPU (central processing unit)- based collision detection. Hence, systems of millions of particles interacting with complex dynamic systems can be efficiently analyzed, and design recommendations can be made in a much shorter time. The figure shows an example of this capability where the Brazil Nut problem is simulated: as the container full of granular material is vibrated, the large ball slowly moves upwards. This capability was expanded to account for anchors of different shapes and penetration velocities, interacting with granular soils.
Quadrelli, Marco B.↗
Characterizing the performance of node-aware strategies for irregular point-to-point communication on heterogeneous architectures
Supercomputer architectures are trending toward higher computational throughput due to the inclusion of heterogeneous compute nodes. These multi-GPU nodes increase on-node computational efficiency, while also increasing the amount of data to be communicated and the number of potential data flow paths. In this work, we characterize the performance of irregular point-to-point communication with MPI on heterogeneous compute environments through performance modeling, demonstrating the limitations of standard communication strategies for both device-aware and staging-through-host communication techniques. Presented models suggest staging communicated data through host processes then using node-aware communication strategies for high inter-node message counts. Notably, the models also predict that node-aware communication utilizing all available CPU cores to communicate inter-node data leads to the most performant strategy when communicating with a high number of nodes. Furthermore, model validation is provided via a case study of irregular point-to-point communication patterns in distributed sparse matrix–vector products. Importantly, we include a discussion on the implications model predictions have on communication strategy design for emerging supercomputer architectures.
97 MATHEMATICS AND COMPUTING↗
Performance Evaluation of Heterogeneous GPU Programming Frameworks for Hemodynamic Simulations
Preparing for the deployment of large scientific and engineering codes on upcoming exascale systems with GPU-dense nodes is made challenging by the unprecedented diversity of device architectures and heterogeneous programming models. In this work, we evaluate the process of porting a massively parallel, fluid dynamics code written in CUDA to SYCL, HIP, and Kokkos with a range of backends, using a combination of automated tools and manual tuning. We use a proxy application along with a custom performance model to inform the results and identify additional optimization strategies. At scale performance of the programming model implementations are evaluated on pre-production GPU node architectures for Frontier and Aurora, as well as on current NVIDIA device-based systems Summit and Polaris. Real-world workloads representing 3D blood flow calculations in complex vasculature are assessed. Our analysis highlights critical trade-offs between code performance, portability, and development time.
Martin, Aristotle↗
BMX: Biological modelling and interface exchange
Abstract High performance computing has a great potential to provide a range of significant benefits for investigating biological systems. These systems often present large modelling problems with many coupled subsystems, such as when studying colonies of bacteria cells. The aim to understand cell colonies has generated substantial interest as they can have strong economic and societal impacts through their roles in in industrial bioreactors and complex community structures, called biofilms, found in clinical settings. Investigating these communities through realistic models can rapidly exceed the capabilities of current serial software. Here, we introduce BMX, a software system developed for the high performance modelling of large cell communities by utilising GPU acceleration. BMX builds upon the AMRex adaptive mesh refinement package to efficiently model cell colony formation under realistic laboratory conditions. Using simple test scenarios with varying nutrient availability, we show that BMX is capable of correctly reproducing observed behavior of bacterial colonies on realistic time scales demonstrating a potential application of high performance computing to colony modelling. The open source software is available from the zenodo repository https://doi.org/10.5281/zenodo.8084270 under the BSD-2-Clause licence.
97 MATHEMATICS AND COMPUTING↗
Initial comparison of single cylinder Stirling engine computer model predictions with test results
A NASA developed digital computer code for a Stirling engine, modelling the performance of a single cylinder rhombic drive ground performance unit (GPU), is presented and its predictions are compared to test results. The GPU engine incorporates eight regenerator/cooler units and the engine working space is modelled by thirteen control volumes. The model calculates indicated power and efficiency for a given engine speed, mean pressure, heater and expansion space metal temperatures and cooler water inlet temperature and flow rate. Comparison of predicted and observed powers implies that the reference pressure drop calculations underestimate actual pressure drop, possibly due to oil contamination in the regenerator/cooler units, methane contamination in the working gas or the underestimation of mechanical loss. For a working gas of hydrogen, the predicted values of brake power are from 0 to 6% higher than experimental values, and brake efficiency is 6 to 16% higher, while for helium the predicted brake power and efficiency are 2 to 15% higher than the experimental.
Tew, R. C., Jr.↗
ERF: Energy Research and Forecasting Model
High performance computing (HPC) architectures have undergone rapid development in recent years. As a result, established software suites face an ever increasing challenge to remain performant on and portable across modern systems. Many of the widely adopted atmospheric modeling codes cannot fully (or in some cases, at all) leverage the acceleration provided by General-Purpose Graphics Processing Units, leaving users of those codes constrained to increasingly limited HPC resources. Energy Research and Forecasting (ERF) is a regional atmospheric modeling code that leverages the latest HPC architectures, whether composed of only Central Processing Units (CPUs) or incorporating GPUs. ERF contains many of the standard discretizations and basic features needed to model general atmospheric dynamics. The modular design of ERF provides a flexible platform for exploring different physics parameterizations and numerical strategies. ERF is built on a state-of-the-art, well-supported, software framework (AMReX) that provides a performance portable interface and ensures ERF's long-term sustainability on next generation computing systems. This paper details the numerical methodology of ERF, presents results for a series of verification/validation cases, and documents ERF's performance on current HPC systems. The roughly 5× speed up of ERF (using GPUs) over Weather Research and Forecasting (CPUs only) for a 3D squall line test case highlights the significance of leveraging GPU acceleration.
17 WIND ENERGY↗
Celeritas Midterm SciDAC Report
Celeritas is a new Monte Carlo (MC) code that helps satisfy the increasing demand for high energy physics (HEP) detector simulation, using Graphics Processing Unit (GPU) hardware on high performance computing (HPC) systems to model Large Hadron Collider (LHC) experiments and beyond. This report details the project’s progress midway through its SciDAC funding period, highlighting the first complete implementation of standard electromagnetic (EM) physics on GPUs, initial results for performance and scalability on Leadership Computing Facilities (LCFs), and preliminary integration into the CMS and ATLAS experiments. By integrating HEP domain knowledge with expertise in MC transport, Celeritas has catalyzed a shift in the HEP community’s perception of GPU platforms as the future for HPC simulations.
97 MATHEMATICS AND COMPUTING↗
A high‐order discontinuous Galerkin approach for physics‐based thermospheric modeling
Abstract The accurate prediction of aerodynamic drag on satellites orbiting in the upper atmosphere is critical to the operational success of modern space technologies, such as satellite‐based communication or navigation systems, which have become increasingly popular in the last few years due to the deployment of constellations of satellites in low‐Earth orbit. As a result, physics‐based models of the ionosphere and thermosphere have emerged as a necessary tool for the prediction of atmospheric outputs under highly variable space weather conditions. This paper proposes a high‐fidelity approach for physics‐based space weather modeling based on the solution of the Navier–Stokes equations using a high‐order discontinuous Galerkin method, combined with a matrix‐free strategy suitable for high‐performance computing on GPU architectures. The approach consists of a thermospheric model that describes a chemically frozen neutral atmosphere in nonhydrostatic equilibrium driven by the external excitation of the Sun. A novel set of variables is considered to treat the low densities present in the upper atmosphere and to accommodate the wide range of scales present in the problem. At the same time, and unlike most existing approaches, radial and angular directions are treated in a nonsegregated approach. The study presents a set of numerical examples that demonstrate the accuracy of the approximation and validate the current approach against observational data along a satellite orbit, including estimates of established empirical and physics‐based models of the ionosphere‐thermosphere system. Finally, a one‐dimensional radial derivation of the physics‐based model is presented and utilized for conducting a parametric study of the main thermal quantities under various solar conditions.
Engineering↗
Hybrid programming-model strategies for GPU offloading of electronic structure calculation kernels
To address the challenge of performance portability and facilitate the implementation of electronic structure solvers, we developed the basic matrix library (BML) and Parallel, Rapid O(N), and Graph-based Recursive Electronic Structure Solver (PROGRESS) library. The BML implements linear algebra operations necessary for electronic structure kernels using a unified user interface for various matrix formats (dense and sparse) and architectures (CPUs and GPUs). Focusing on density functional theory and tight-binding models, PROGRESS implements several solvers for computing the single-particle density matrix and relies on BML. In this paper, we describe the general strategies used for these implementations on various computer architectures, using OpenMP target functionalities on GPUs, in conjunction with third-party libraries to handle performance critical numerical kernels. In this study, we demonstrate the portability of this approach and its performance in benchmark problems.
36 MATERIALS SCIENCE↗
Toward Energy-Efficient HPC: Insights from Power Profiling a Cloud-Resolving Earth System Model
Power is a fundamental constraint as supercomputing advances to exascale. Efficient operation within strict power budgets requires application-aware power management based on a detailed understanding of application-level power behavior. This work analyzes the Energy Exascale Earth System Model (E3SM) atmosphere component, SCREAM, on Perlmutter (NERSC) and Frontier (OLCF). We characterize power variation across inputs, concurrency levels, and power caps, evaluate the energy impact of code optimizations, and attribute energy within the code using a newly developed GPU energy model. Results show that SCREAM’s peak power remains stable during its core execution phase and decreases gradually as concurrency increases. Power capping experiments reveal a performance–energy "sweet spot". On Perlmutter, limiting GPU power to 50% of thermal design power (TDP) achieves up to 15% energy savings with a 7% performance penalty. On Frontier, a 40% TDP cap yields up to 10% energy savings with less than 10% performance loss. Code optimizations reduce SCREAM energy by shortening run time without increasing power. Modeling reveals a critical insight: data movement accounts for approximately 70% of SCREAM’s GPU energy. This fundamentally shifts the optimization focus from FLOPS to data transfer reduction for this class of applications, offering the most impactful strategy for improving energy efficiency. This work establishes a foundation for practical, application-aware power management at exascale.
Zhao, Zhengji [Lawrence Berkeley National Laborato↗