Search NASA⌕ Search

SEARCH · Search NASA

Results for “data usage”

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 145 records · Page 8
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Human Factors and Technologies Design to Improve User Acceptance of Pooled Rideshare for Increasing Transportation System Energy Efficiency

This multi-year project delivered a comprehensive, human-factors-driven framework to understand, model, and improve pooled rideshare (PR) adoption in the United States. Through three large-scale national survey studies involving more than 16,000 participants across multiple cities and demographic groups, the research established one of the most extensive datasets to date on user perceptions, behavioral barriers, and service expectations related to pooled rideshare. These data revealed key human factors barriers of user acceptance of PR and suggested potential actionable experience optimizations that could lead to increased PR usage. This foundational knowledge guided the development of novel human-factors models and behavioral choice models that quantify how psychological, demographic, and trip-level factors influence willingness to pool. Building on these empirical insights, the project developed advanced behavioral modeling tools, including mixed logit and integrated choice and latent variable models, to capture both observable and latent influences on PR adoption. These models significantly improved the ability to predict riders’ acceptance of pooled trips, explaining choice heterogeneity through latent constructs such as safety, service experience, privacy concerns, time sensitivity, and environmental attitudes. Together, these models provide a robust analytical foundation for designing PR systems that more effectively meet user needs. The project translated human-factors insights and behavioral models into actionable technology innovations by extending POLARIS—an agent-based, activity-based travel simulation platform—into a fully functional pooled rideshare simulation environment. New PR modules, acceptance models, and regional scenarios were implemented for Greenville, SC and Austin, TX, enabling high-fidelity validation of algorithmic strategies under realistic demand and traffic conditions. The simulation platform supported the development and evaluation of adaptive discount-based assignment algorithms, enhanced willingness-to-pay formulations, demographic-aware incentive mechanisms, and a proactive joint assignment and repositioning strategy. Simulation results demonstrated substantial gains in pooling uptake, average vehicle occupancy, energy efficiency, and fleet profitability. In Greenville, pooling adoption more than doubled, while reductions in vehicle-miles traveled and energy consumption were significant. In Austin, pooling improvements were achieved with minimal service-quality trade-offs, and profitability increased across all fleet sizes. Through this research, we developed a comprehensive understanding of the human factors barriers that limit user acceptance of pooled rideshare services. These insights enabled the design of human-factors-aware pooled rideshare technologies that more effectively address user concerns and improve adoption rates. By integrating these models into an advanced agent-based simulation framework, we demonstrated that higher adoption of pooled rideshare can lead to measurable improvements in energy efficiency and system performance. Together, these contributions establish a validated pathway from human-centered analysis to technology development and energy-saving outcomes, supporting national goals for more sustainable and efficient mobility systems.

Jia, Yunyi↗
DOE OSTI2025

rcsb-api : Python Toolkit for Streamlining Access to RCSB Protein Data Bank APIs

The Protein Data Bank (PDB) was founded in 1971 as the first open-access digital data resource in biology to serve as the single global archive for three-dimensional (3D) macromolecular structure data. Current PDB holdings exceed 230,000 experimentally determined structures of proteins, nucleic acids, viruses, and macromolecular machines. The RCSB Protein Data Bank RCSB.org research-focused web portal facilitates search, analyses, and visualization of every PDB structure along with more than one million Computed Structure Models from AlphaFold DB and the ModelArchive. It is powered by a set of publicly available Application Programming Interfaces (APIs) that both support RCSB.org users and provide programmatic access to PDB data. Given the breadth and levels of granularity encompassed in this rich data collection, efficiently accessing the information programmatically may be challenging for new users. RCSB PDB has developed a Python software package, rcsb-api , that facilitates easy and efficient use of RCSB PDB APIs within a Python environment. This software tool is designed to streamline access to the extensive corpus of data housed within the PDB, enabling researchers to search, retrieve, and analyze 3D biostructure data seamlessly. Its use will accelerate research in structural biology, molecular biology and biochemistry, drug discovery, and bioinformatics by providing more efficient tools for data integration and analysis. The new toolkit is available on GitHub (github.com/rcsb/py-rcsb-api) and published to the public Python package repository (PyPI) to foster wider usage and support basic and applied research in fundamental biology, biomedicine, and the energy sciences.

FAIR principles↗
DOE OSTI2026

SpaceNet 9—Cross-Sensor Alignment of Optical and SAR Imagery

Precise registration of high-resolution synthetic aperture radar (SAR) and optical imagery is necessary for realizing the full potential and benefits of multimodal image analysis. However, two significant challenges presently exist. First, there is a lack of annotated datasets and benchmarks available for high-resolution SAR–optical image registration. Second, an assessment of efficient and reliable image registration methods that can precisely align these modalities is lacking. Here, we present a holistic description of the SpaceNet 9 Challenge and its results. We present a description of the dataset and baseline algorithm along with the results of the challenge, including a description of the winning algorithms. We release the SpaceNet 9 dataset along with open-sourcing the winning algorithms and baseline. The objective of SpaceNet 9 was to compute a dense displacement map that indicates the shift needed to align pixels in an optical image to the pixels in a SAR image. The challenge launched in April 2025 and was active for approximately two months. The top five solutions reduced image alignment error from approximately 34 m to under 13 m for public and private test data, with the best results obtaining a registration error of only 8.5 and 6.7 m on the public testing and private testing dataset, respectively. Usage of pretrained image matching models, robust outlier rejection with RANSAC, and estimating local displacement were common among the top solutions. The results of this challenge provide insight into high-resolution SAR–optical image registration and offer opportunities for future benchmarking in this domain. The baseline algorithm, winning solutions, and datasets are available at https://spacenet.ai/sn9-challenge/.

benchmark datasets↗
DOE OSTI2026

Transit Rider/Travel Behavior Inventory Survey - Minneapolis-St. Paul Metro - 2010

The Metropolitan Council administered a system-wide transit on-board survey between September and November 2010 as part of the 2010 Travel Behavior Inventory to provide detailed transit usage patterns and rider information to support modeling and planning efforts. The study’s focus was to capture only the most current and reliable data necessary to determine future public transportation needs in the Minneapolis region. The study was structured to collect detailed transit ridership data for different routes during different times of day to develop a disaggregate transit trip table that will support advanced travel demand modeling. The study was designed to leverage existing data sources, such as the 2005 on-board survey, to provide the best quality data to update the Travel Behavior Inventory in the Minneapolis-Saint Paul metropolitan area.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

Characterization of a CMOS camera based film digitization platform for gated x-ray imaging diagnostics at the National Ignition Facility

Hardened gated x-ray detectors use photographic film as the data recording medium due to its low sensitivity to the high-yield neutron environments at the National Ignition Facility (NIF). The photographic film is digitized with a Photometric Data Systems (PDS) microdensitometer, which measures the film’s optical density. The PDS scanner is able to measure a dynamic range of 0–5 OD; however, raster scanning the film is time consuming and maintenance of the instrument is challenging due to legacy technology. Since film usage at NIF is expected to continue in the foreseeable future, a digitization platform that is faster and more maintainable would benefit the NIF’s current and future operations. Here, this work presents the characterization of the digital transitions (DT) atom, a CMOS camera-based digitization platform that records film data in a single image capture very quickly and has widely available user support. The preliminary results suggest that the DT atom is able to reconstruct exposures accurately enough to be a competitive alternative to the PDS Scanner.

47 OTHER INSTRUMENTATION↗
DOE OSTI2024

Generic and ML Workloads in an HPC Datacenter: Node Energy, Job Failures, and Node-Job Analysis

HPC datacenters offer a backbone to the modern digital society. Increasingly, they run Machine Learning (ML) jobs next to generic, compute-intensive workloads, supporting science, business, and other decision-making processes. However, understanding how ML jobs impact the operation of HPC datacenters, relative to generic jobs, remains desirable but understudied. In this work, we leverage long-term operational data, collected from a national-scale production HPC datacenter, and statistically compare how ML and generic jobs can impact the performance, failures, resource utilization, and energy consumption of HPC datacenters. Our study provides key insights, e.g., ML-related power usage causes GPU nodes to run into temperature limitations, median/mean runtime and failure rates are higher for ML jobs than for generic jobs, both ML and generic jobs exhibit highly variable arrival processes and resource demands, significant amounts of energy are spent on unsuccessfully terminating jobs, and concurrent jobs tend to terminate in the same state. We open-source our cleaned-up data traces on Zenodo (https://doi. org/10.5281/zenodo.13685426), and provide our analysis toolkit as software hosted on GitHub (https://github.com/atlarge-research/2024-icpads-hpc-workload-characterization). This study offers multiple benefits for data center administrators, who can improve operational efficiency, and for researchers, who can further improve system designs, scheduling techniques, etc.

crossanalysis↗
DOE OSTI2024

Field Study of 120-volt Heat Pump Water Heaters in the Big Easy

To meet long-term goals for reducing carbon emissions and lessen the impacts of climate change, the U.S. plans to decarbonize its building stock, which includes the replacement of fossil fuel-burning end uses with energy efficient electric alternatives. As part of this strategy, the replacement of fossil fuel water heaters with heat pump water heaters (HPWH) has the potential to avoid substantial carbon emissions. An estimated 10-15 million single-family homes with fossil fuel water heaters do not have the electrical panel capacity to install a 240-volt, 30-amp HPWH. For these homes, a technology recently introduced to the market, the 120-volt plug-in HPWH, can provide energy efficient electrification of water heating without an expensive panel and wiring retrofit. This paper presents the results of an ongoing 120-volt HPWH field study conducted in 17 homes in New Orleans, LA. Key installation scenarios are profiled for retrofitting from gas-fired water heaters to 120-volt HPWHs, accounting for space, air volume, air temperature, condensate drainage, and electrical. Each HPWH had its surrounding air temperature, relative humidity, inlet and outlet water temperature, flow, and energy consumption monitored. Using these data, hot water delivery and energy efficiency performance were analyzed based on home characteristics. Hot water run outs were investigated to understand how hot water usage, inlet water temperature, and surrounding air temperature impact the HPWH’s ability to meet load. From these results, best practices for 120-volt HPWH siting, sizing, and installation were developed. In addition, results are explored further to add insight for electrification policies and product development.

heat pump water heater, 120-volt, Residential buil↗
DOE OSTI2026

Grid Operator Analytics and Assessment Tools for Inverter- Based Resources Dominated Grid (GOAAT-IBR) Project Update

This presentation provides an update on the OPTIMA GOAAT project, with emphasis on the cloud-native data platform developed in-house to ingest, manage, and operationalize high-resolution power system data. Since our last NASPI presentation, accessible via OSTI ID #2671437, the project team advanced the design and deployment of a scalable architecture capable of handling both synchronized and non-synchronized streams, including PMU, point-on-wave (POW), COMTRADE, and SCADA data. These materials review the project status, recent progress, and key lessons learned. The core of the presentation examines the architecture and engineering of our cloud-native ingestion and data management platform. We then explain how pipelines were designed to collect, normalize, time-align, store, and serve heterogeneous data at scale. We will discuss design choices such as data models, streaming versus batch ingestion, storage tiers, and interoperability with analytics applications. Practical experiences with cloud-native technologies were shared during the event, including benefits, limitations, and integration challenges in a utility environment, along with methods used to improve performance, reduce latency, and optimize resource usage. The presentation also showcases user interface designs and visualization tools that convert raw measurements and analytics results into intuitive, actionable insights for operators and engineers. During the presentation examples were provided demonstrating how visualization, event views, and summarized analytics enhance situational awareness and support operational decision-making. These use cases illustrate how a well-designed data infrastructure can bridge the gap between high-volume measurements and practical grid operations.

Aminifar, Farrokh↗
DOE OSTI2024

High-throughput genetics enables identification of nutrient utilization and accessory energy metabolism genes in a model methanogen

Archaea are widespread in the environment and play fundamental roles in diverse ecosystems; however, characterization of their unique biology requires advanced tools. This is particularly challenging when characterizing gene function. Here, we generate randomly barcoded transposon libraries in the model methanogenic archaeon Methanococcus maripaludis and use high-throughput growth methods to conduct fitness assays (RB-TnSeq) across over 100 unique growth conditions. Using our approach, we identified new genes involved in nutrient utilization and response to oxidative stress. We identified novel genes for the usage of diverse nitrogen sources in M. maripaludis including a putative regulator of alanine deamination and molybdate transporters important for nitrogen fixation. Furthermore, leveraging the fitness data, we inferred that M. maripaludis can utilize additional nitrogen sources including $\tiny{L}$-glutamine, $\tiny{D}$-glucuronamide, and adenosine. Under autotrophic growth conditions, we identified a gene encoding a domain of unknown function (DUF166) that is important for fitness and hypothesize that it has an accessory role in carbon dioxide assimilation. Finally, comparing fitness costs of oxygen versus sulfite stress, we identified a previously uncharacterized class of dissimilatory sulfite reductase-like proteins (Dsr-LP; group IIId) that is important during growth in the presence of sulfite. When overexpressed, Dsr-LP conferred sulfite resistance and enabled use of sulfite as the sole sulfur source. The high-throughput approach employed here allowed for generation of a large-scale data set that can be used as a resource to further understand gene function and metabolism in the archaeal domain.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

$\bar{\nu}_\mu$ charged-current $\pi^0$ data release

Data release for the NOvA muon antineutrino charged-current (CC) pi^0 cross section presented in arXiv:2511.05807. The signal for this analysis is defined as muon antineutrino CC interactions in the fiducial volume of the NOvA near detector (a 2.7 m × 2.7 m × 9.0 m region) that produce at least one pi^0 in the final state emerging from the nucleus, within the phase space of muon momentum [0.5, 2.5) GeV/c and muon angle [0, 60) degree, as described in arXiv:2511.05807. The released zip file contains two files: NOvA_NumubarCCPi0_DataRelease.root README.txt The ROOT file includes the cross-section results as well as the statistical and systematic covariance matrices for each variable used in this analysis. The README provides a detailed description of the contents of the data release. Official Flux: The flux used in this analysis is available from the NOvA Public Docs: https://publicdocs.fnal.gov/cgi-bin/ShowDocument?docid=8. File structure --- The ROOT file contains the following TDirectories: pi0p - pi^0 momentum distributions pi0dir - pi^0 angular distributions muonp - muon momentum distributions muondir - muon angular distributions Q2 - reconstructed Q^2 distributions Wmass - reconstructed W_mass distributions Each directory contains three histograms: xsec (TH1D): Cross section result cov_stat (TH2D): Statistical covariance matrix cov_syst (TH2D): Systematic covariance matrix Usage notes: - xsec gives the measured differential cross section w.r.t. the corresponding variable. - cov_stat and cov_syst provide the full covariance matrices. - The bin definitions and kinematic phase spaces follow those used in arXiv:2511.05807. Citation --- If you use these data, please cite: NOvA Collaboration, arXiv:2511.05807.

Wu, Wanwei [Pittsburgh U.] (ORCID:0000000326327215↗
DOE OSTI2024

North America’s Potential for an Environmentally Sustainable Nickel, Manganese, and Cobalt Battery Value Chain

The Detroit Big Three General Motors (GMs), Ford, and Stellantis predict that electric vehicle (EV) sales will comprise 40–50% of the annual vehicle sales by 2030. Among the key components of LIBs, the LiNixMnyCo1−x−yO2 cathode, which comprises nickel, manganese, and cobalt (NMC) in various stoichiometric ratios, is widely used in EV batteries. This review reveals NMC cathodes from laboratory research. Furthermore, this study examines the environmental effect of NMC cathode production for EV batteries (including coating technologies), encompassing aspects such as energy consumption, water usage, and air emissions. Although gaps persist in NMC cathode environmental assessments (NMC111, NMC532, NMC622, and NMC811), limited life cycle assessments “(LCA)” have been conducted. Most available data originate from Asia (primarily China), accounting for 85% of the production of EV LIB cathode materials. The concept of battery passports for data collection on LIB components has been proposed to facilitate material traceability as a system for ensuring a sustainable supply chain for critical minerals. The automotive industry’s shift to electrification necessitates a sustainable supply chain from mine to vehicle end-of-life. As the critical mineral supply moves from Asia to North America, environmentally friendly industrial methods must be studied to provide this supply chain direction.

25 ENERGY STORAGE↗
DOE OSTI2025

Hardware acceleration for HPS algorithms in two and three dimensions

We provide a flexible, open-source framework for hardware acceleration, namely massively-parallel execution on general-purpose graphics processing units (GPUs), applied to the hierarchical Poincaré–Steklov (HPS) family of algorithms for building fast direct solvers for linear elliptic partial differential equations. To take full advantage of the power of hardware acceleration, we propose two variants of HPS algorithms to improve performance on two- and three-dimensional problems. In the two-dimensional setting, we introduce a novel recomputation strategy that minimizes costly data transfers to and from the GPU; in three dimensions, we modify and extend the adaptive discretization technique of Geldermans and Gillman [1] to greatly reduce peak memory usage. We provide an open-source implementation of these methods written in JAX, a high-level accelerated linear algebra package, which allows for the first integration of a high-order fast direct solver with automatic differentiation tools. We conclude with extensive numerical examples showing our methods are fast and accurate on two- and three-dimensional problems.

Fast direct solvers↗
DOE OSTI2024

Power Profile Monitoring and Tracking Evolution of System-Wide HPC Workloads

The power & energy demands of HPC machines have grown significantly. Modern exascale HPC systems require tens of megawatts of combined power for computing resources and cooling facilities at full capacity. The current energy trend is not sustainable for future HPC systems, and there is a need to work toward the energy efficiency aspect of HPC performance. Energy awareness of the HPC applications at the job level is essential for running an efficient HPC system. This work aims to develop a pipeline to provide a production-level system-wide overview of the HPC workloads' power profile while handling evolving workloads exhibiting new power trends. We developed an open-set classification model for HPC jobs based on the properties of power profiles to continuously provide a system-wide holistic view of recently completed jobs. The pipeline helps continuously monitor the job-level power usage pattern of HPC and enables us to capture the new trends in applications' power behavior. We employed a comprehensive set of techniques to generate job-level data, custom-designed feature extraction methods to extract critical features from jobs' power profiles, clustering techniques powered by generative modeling, and open-set classification for identifying job profiles into known classes or an unknown set. With extensive evaluations, we demonstrate the effectiveness of each component in our pipeline. We provide an analysis of the resulting clusters that characterize the power profile landscape of the Summit supercomputer from more than 60K jobs executed in a year. The open-set classification classifies the known data sets into known classes with high accuracy and identifies unknown data noints with over 85% accuracy.

Karimi, Ahmad Maroof↗
DOE OSTI2024

RAP: Resource-aware Automated GPU Sharing for Multi-GPU Recommendation Model Training and Input Preprocessing

Ensuring high-quality recommendations for newly onboarded users requires the continuous retraining of Deep Learning Recommendation Models (DLRMs) with freshly generated data. To serve the online DLRM retraining, existing solutions use hundreds of CPU computing nodes designated for input preprocessing, causing significant power consumption that surpasses even the power usage of GPU trainers. To this end, we propose RAP, an end-to-end DLRM training framework that supports Resource-aware Automated GPU sharing for DLRM input Preprocessing and Training. The core idea of RAP is to accurately capture the remaining GPU computing resources during DLRM training for input preprocessing, achieving superior training efficiency without requiring additional resources. Specifically, RAP utilizes a co-running cost model to efficiently assess the costs of various input preprocessing operations, and it implements a resource-aware horizontal fusion technique that adaptively merges smaller kernels according to GPU availability, circumventing any interference with DLRM training. In addition, RAP leverages a heuristic searching algorithm that jointly optimizes both the input preprocessing graph mapping and the co-running schedule to maximize the end-to-end DLRM training throughput. The comprehensive evaluation shows that RAP achieves 78.3× speedup on average over CPU-based DLRM input preprocessing frameworks. In addition, the end-to-end training throughput of RAP is only 2.04% lower than the ideal case, which has no input preprocessing overhead.

Wang, Zheng↗
DOE OSTI2024

Platform for Remote Deployment and Training for Enhanced Building Operation Practices (Building Re-Tuning and On-going Commissioning)

While a building’s energy usage is driven largely by its design and use, building operator behavior has a strong influence on its energy consumption. This project developed and piloted a specific, data-driven coaching methodology to help operators understand how they can adjust operations and/or affect no/low-cost repairs or upgrades to their specific building HVAC systems to reduce energy consumption. Named BuildingCoach, the operational optimization method used is based on the Building Re-tuning approach developed by the Pacific Northwest National Laboratory. A building operations analytics market has matured over the past decade, though its potential to affect energy-saving changes has not been fully realized. Training operators to understand the methods for operational optimization with the explicit approach of using building-system performance data is hypothesized to create a more effective, longer lasting result in building energy efficiency, and this strategy is the fundamental premise of this project. With the support of an Industry Advisory Board, the project succeeded in developing materials and recruiting for and delivering three pilot cohorts. Deliverables included twenty-two self-paced training modules (accessed via a Learning Management System) and a web-based platform that includes access to real-time building system data and a repository for building system documentation. The project set out to have 100 participants from 50 buildings in three pilot cohorts. In the end, there were 28 participants from 17 buildings, i.e., a significant shortfall. The first two pilot cohorts had only two buildings in each, and this was partially due to difficulties in deploying the Building Operator Coaching Solution (“the BOCS”), which is technology that extracts the data from the controls network and presents it as prescribed for coaching. In the third cohort, the project team deployed the BOCS successfully to 13 buildings, the methodology was piloted as intended, and numerous opportunities for optimization were identified. The BuildingCoach business plan charts a path to an economically sustainable effort. However, even with a licensing model captured in the final version of the business plan, the scalability is still limited to keeping less than 1,000 buildings affected by 2033. Even so, there are unexplored paths to greater scalability that are being considered. CUNY BPL is working to perpetuate and grow the use of BuildingCoach. As of this writing, about twenty buildings have either been connected or will be connected with operators coached / to be coached in the NYC municipal portfolio, twelve buildings across four campuses in NY State will use BuildingCoach, a NY upstate county wishes for six or seven buildings to participate with the support of funding from NYSERDA, and others have also expressed interest. In the decades to come, there will be an increasing percentage of large and mid-sized buildings that incorporate automated system optimization (ASO), and the building operators’ role will shift to spend more time on maintenance and monitoring. Meanwhile, programs such as BuildingCoach will play a critical role in optimizing operations. And, regardless of the emergence of ASO, operators will still need to understand how their systems operate so that they can monitor them properly. Within that context, BuildingCoach is an important step towards operators’ understanding of efficient building system operations.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

Accelerating science: The usage of commercial clouds in ATLAS Distributed Computing

The ATLAS experiment at CERN is one of the largest scientific machines built to date and will have ever growing computing needs as the Large Hadron Collider collects an increasingly larger volume of data over the next 20 years. ATLAS is conducting R&D projects on Amazon Web Services and Google Cloud as complementary resources for distributed computing, focusing on some of the key features of commercial clouds: lightweight operation, elasticity and availability of multiple chip architectures. The proof of concept phases have concluded with the cloud-native, vendoragnostic integration with the experiment’s data and workload management frameworks. Google Cloud has been used to evaluate elastic batch computing, ramping up ephemeral clusters of up to O(100k) cores to process tasks requiring quick turnaround. Amazon Web Services has been exploited for the successful physics validation of the Athena simulation software on ARM processors. We have also set up an interactive facility for physics analysis allowing endusers to spin up private, on-demand clusters for parallel computing with up to 4 000 cores, or run GPU enabled notebooks and jobs for machine learning applications. The success of the proof of concept phases has led to the extension of the Google Cloud project, where ATLAS will study the total cost of ownership of a production cloud site during 15 months with 10k cores on average, fully integrated with distributed grid computing resources and continue the R&D projects.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Speeding up and reducing memory usage for scientific machine learning via mixed precision

Scientific machine learning (SciML) has emerged as a versatile approach to address complex computational science and engineering problems. Within this field, physics-informed neural networks (PINNs) and deep operator networks (DeepONets) stand out as the leading techniques for solving partial differential equations by incorporating both physical equations and experimental data. However, training PINNs and DeepONets require significant computational resources, including long computational times and large amounts of memory. In search of computational efficiency, training neural networks using half precision (float16) rather than the conventional single (float32) or double (float64) precision has gained substantial interest, given the inherent benefits of reduced computational time and memory consumed. However, we find that float16 cannot be applied to SciML methods, because of gradient divergence at the start of training, weight updates going to zero, and the inability to converge to a local minima. To overcome these limitations, we explore mixed precision, which is an approach that combines the float16 and float32 numerical formats to reduce memory usage and increase computational speed. Our experiments showcase that mixed precision training not only substantially decreases training times and memory demands but also maintains model accuracy. Here, we also reinforce our empirical observations with a theoretical analysis. The research has broad implications for SciML in various computational applications.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Using the ATLAS experiment software on heterogeneous resources

With the large dataset expected from 2030 onwards by the HL-LHC at CERN, the ATLAS experiment is reaching the limits of the current data processing model in terms of traditional CPU resources based on x86_64 architectures and an extensive program for software upgrades towards the HL-LHC has been set up. The ARM CPU architecture is becoming a competitive and energy efficient alternative. Accelerators like GPUs are available in any recent HPC. In the past years ATLAS has successfully ported its full data processing and simulation software framework Athena to ARM and has invested significant effort in porting parts of the reconstruction and simulation algorithms to GPUs. We report on the successful usage of the ATLAS experiment offline and online software framework Athena on ARM and GPUs through the PanDA workflow management system at various WLCG sites. Furthermore we report on performance optimizations of the builds for ARM CPUs and the GPU integration efforts. We will discuss performance comparisons of different ARM and x86_64 architectures on WLCG resources and Cloud compute providers like GCP and AWS using ATLAS productions workflows as used in the Hep-Score23 benchmark suite.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗