SEARCH · Engineering Papers

Results for “AI hardware”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 199 records · Page 11

All sources NASA NTRS DOE OSTI

DOE OSTI2023

Mixed-Precision S/DGEMM Using the TF32 and TF64 Frameworks on Low-Precision AI Tensor Cores

Using NVIDIA graphics processing units (GPUs) equipped with Tensor Cores has enabled the significant acceleration of general matrix multiplication (GEMM) for applications in machine learning (ML) and artificial intelligence (AI) and in high-performance computing (HPC) generally. The use of such power-efficient, specialized accelerators can provide a performance increase between 8 × and 20 ×, albeit with a loss in precision. However, a high level of precision is required in many large scientific and HPC applications, and computing in single or double precision is still necessary for many of these applications to maintain accuracy. Fortunately, mixed-precision methods can be employed to maintain a higher level of numerical precision while also taking advantage of the performance increases from computing with lower-precision AI cores. With this in mind, we extend the state of the art by using NVIDIA’s new TF32 framework. This new framework not only burdens some constraints of the previous frameworks, such as costly 32 16-bit castings but also provides an equivalent precision and performance by using a much simpler approach. We also propose a new framework called TF64 that attempts double-precision arithmetic with low-precision Tensor Cores. Although this framework does not exist yet, we validated the correctness of this idea and achieved an equivalent of 64-bit precision on 32-bit hardware.

Valero Lara, Pedro↗

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↗

NASA NTRS2015

Towards an Open, Distributed Software Architecture for UxS Operations

To address the growing need to evaluate, test, and certify an ever expanding ecosystem of UxS platforms in preparation of cultural integration, NASA Langley Research Center's Autonomy Incubator (AI) has taken on the challenge of developing a software framework in which UxS platforms developed by third parties can be integrated into a single system which provides evaluation and testing, mission planning and operation, and out-of-the-box autonomy and data fusion capabilities. This software framework, named AEON (Autonomous Entity Operations Network), has two main goals. The first goal is the development of a cross-platform, extensible, onboard software system that provides autonomy at the mission execution and course-planning level, a highly configurable data fusion framework sensitive to the platform's available sensor hardware, and plug-and-play compatibility with a wide array of computer systems, sensors, software, and controls hardware. The second goal is the development of a ground control system that acts as a test-bed for integration of the proposed heterogeneous fleet, and allows for complex mission planning, tracking, and debugging capabilities. The ground control system should also be highly extensible and allow plug-and-play interoperability with third party software systems. In order to achieve these goals, this paper proposes an open, distributed software architecture which utilizes at its core the Data Distribution Service (DDS) standards, established by the Object Management Group (OMG), for inter-process communication and data flow. The design decisions proposed herein leverage the advantages of existing robotics software architectures and the DDS standards to develop software that is scalable, high-performance, fault tolerant, modular, and readily interoperable with external platforms and software.

Cross, Charles D.↗

DOE OSTI2023

AI & Physics-Based Bad Command/Data Detection in Large Power Electronics Systems: Multi-Port Autonomous Reconfigurable Solar Power Plant (MARS)

Detection of bad data from measurement sensors and bad commands from control centers need to be carried out to avoid instabilities within large power electronics systems. Towards the same, in this paper, model and data driven methods are proposed to identify anomalies in measured data and commands received by large power electronics systems. The large power electronics system considered in this paper is a multi-port autonomous reconfigurable solar power plant (MARS), which consists of photovoltaic (PV) and energy storage systems (ESSs) that connect to high-voltage direct current (HVdc) system and transmission ac power grid. The proposed algorithms in the MARS power plant to detect bad data from measurements and bad commands from control centers are evaluated in simulations and hardware-in-the-loop (HIL) tests. Furthermore, it has been observed that the proposed algorithms are able to detect bad measurements and commands in all the use cases evaluated.

14 SOLAR ENERGY↗

DOE OSTI2024

Pretraining Billion-Scale Geospatial Foundational Models on Frontier

As AI workloads increase in scope, generalization capability becomes challenging for small task-specific models and their demand for large amounts of labeled training samples increases. On the contrary, Foundation Models (FMs) are trained with internet-scale unlabeled data via self-supervised learning and have been shown to adapt to various tasks with minimal fine-tuning. Although large FMs have demonstrated significant impact in natural language processing and computer vision, efforts toward FMs for geospatial applications have been restricted to smaller size models, as pretraining larger models requires very large computing resources equipped with state-of-the-art hardware accelerators. Current satellite constellations collect 100+TBs of data a day, resulting in images that are billions of pixels and multimodal in nature. Such geospatial data poses unique challenges opening up new opportunities to develop FMs. We investigate billion scale FMs and HPC training profiles for geospatial applications by pretraining on publicly available data. We studied from end-to-end the performance and impact in the solution by scaling the model size. Our larger 3B parameter size model achieves up to 30% improvement in top1 scene classification accuracy when comparing a 100M parameter model. Moreover, we detail performance experiments on the Frontier supercomputer, America's first exascale system, where we study different model and data parallel approaches using PyTorch's Fully Sharded Data Parallel library. Specifically, we study variants of the Vision Transformer architecture (ViT), conducting performance analysis for ViT models with size up to 15B parameters. By discussing throughput and performance bottlenecks under different parallelism configurations, we offer insights on how to leverage such leadership-class HPC resources when developing large models for geospatial imagery applications.

Tsaris, Aristeidis (aris)↗

NASA NTRSDate not supplied

Enabling Interoperability in Earth System Digital Twins (ESDT): Integrating Observations, Models, and AI for Actionable Insights Through NASA'S Intelligent Systems Technology Program

NASA’s Intelligent Systems Technology Program (IST) is driving a paradigm shift in Earth science through the development of Earth System Digital Twins (ESDT). These integrated information systems create a dynamic "digital replica" of the Earth by harmonizing continuous, multi-source observations with high-fidelity models and state-of-the-art artificial intelligence (AI) that enable “What now?”, “What next?”, and “What if?” scenario building. These scenarios are reflected in NASA IST’s series of ESDTs, from the Coastal Zone Digital Twin that integrates complex data on the current state of the Chesapeake Bay to the Terrestrial Environmental Rapid-Replication and Assimilation Hydrometeorological (TerraHydro) AI-based ESDT that forecasts water movement across Earth’s surface, to the Agriculture Land Information System (AgLIS) which can be used to assess optimal planting dates and crop yield estimates. By bridging the gap between vast data archives and actionable insights, these projects enable a system-of-systems approach to understanding complex, interacting Earth processes. This poster will highlight recent innovations and future directions from NASA’s ESDT initiatives: Continuous Data Assimilation & Multi-Source Fusion. A core requirement of the ESDT work is the transition from static models to dynamic "living" replicas. This involves creating frameworks for the continual assimilation of near-real-time data from uncoordinated, heterogeneous sources, including satellite observations and airborne assets, and ground-based Internet of Things (IoT) sensors. These systems link design, operational status, and environmental data, ensuring the digital twin accurately reflects the current state of the physical Earth system. High-Fidelity Hybrid Modeling & Computational Acceleration to enable interactive "what-if" explorations, programs are moving beyond traditional, slow physical solvers by developing fast surrogate machine learning models and Deep Generative Models (DGMs). These hybrid approaches use neural networks to emulate complex physics, such as cloud feedback or ocean dynamics, at a fraction of the original computing cost, often leveraging advanced hardware like Graphics Processing Units (GPUs) to achieve the necessary scale. Federated Ecosystems & Interoperable Frameworks rather than building isolated tools, NASA IST is moving toward federated ESDTs and reusable analytic collaborative frameworks. This theme focuses on interoperability standards and common ontologies that allow specialized digital twins to interact and share data. This system-of-systems architecture supports multi-discipline investigations, such as analyzing how upstream watershed changes impact downstream urban flooding or how wildfire emissions affect regional air quality. By leveraging these advancements, ESDTs empower researchers and decision-makers to conduct real-time analysis and run complex hypothetical scenarios, ultimately improving our understanding of Earth’s evolving systems and informing critical real-world applications.

Earth System↗

NASA NTRSDate not supplied

Enabling Reliable, Fault-Tolerant Autonomous Lunar Habitats with High-Performance Spaceflight Computing

The lunar surface presents unfavorable constraints and harsh living conditions. To address these challenges, autonomous habitats will require complex integrated systems that combine advanced software, high-performance hardware, and cutting-edge sensors to ensure sustainability, safety, and operational efficiency. Consequently, maintaining a sustainable presence on the Moon requires reliable infrastructure and efficient development, precise monitoring, and utilization of resources within a lunar installation. These elements are essential not only to ensure that lunar settlement can be long-term, self-sustaining, and resource-efficient, but also to serve as a foundation for future missions and eventual human habitation on Mars. Humans are not native to the Moon; therefore, our survival and ability to thrive will depend on autonomous systems that can foster safety and resilience through high-availability architectures, graceful degradation, and highly fault-tolerant spaceflight hardware capable of continuing operation during failures. This requires advanced human-rated distributed systems architectures with specialized electronics, scalable capabilities, and an integrated design approach. Unlike current practices focused on short-term missions and regularly maintained components, permanent lunar compute systems must be designed for extended operations beyond mission durations. This paper explores the necessity of transitioning toward fault- tolerant, highly autonomous hardware systems designed for multi-year missions. It also identifies critical subsystems that require high levels of autonomy, supported by radiation-hardened processors and extreme thermal loads, which are essential to mitigate long-term degradation and ensure sustainable lunar habitation. Finally, the paper aligns with NASA’s identified Civil Space Shortfalls, particularly in high-performance onboard computing, advanced data acquisition, extreme-environment avionics, radiation monitoring and countermeasures, and autonomous health management. It proposes NASA’s new High-Performance Spaceflight Computing (HPSC) processor as a turnkey solution, delivering 100 times the performance-per-watt of legacy rad-hard CPUs and enabling onboard AI, edge computing, and fault-tolerant features essential for sustained lunar autonomy and beyond.

Sarkis S Mikaelian↗

NASA NTRS1993

In search of meta-knowledge

Development of an Intelligent Information System (IIS) involves application of numerous artificial intelligence (AI) paradigms and advanced technologies. The National Aeronautics and Space Administration (NASA) is interested in an IIS that can automatically collect, classify, store and retrieve data, as well as develop, manipulate and restructure knowledge regarding the data and its application (Campbell et al., 1987, p.3). This interest stems in part from a NASA initiative in support of the interagency Global Change Research program. NASA's space data problems are so large and varied that scientific researchers will find it almost impossible to access the most suitable information from a software system if meta-information (metadata and meta-knowledge) is not embedded in that system. Even if more, faster, larger hardware is used, new innovative software systems will be required to organize, link, maintain, and properly archive the Earth Observing System (EOS) data that is to be stored and distributed by the EOS Data and Information System (EOSDIS) (Dozier, 1990). Although efforts are being made to specify the metadata that will be used in EOSDIS, meta-knowledge specification issues are not clear. With the expectation that EOSDIS might evolve into an IIS, this paper presents certain ideas on the concept of meta-knowledge and demonstrates how meta-knowledge might be represented in a pixel classification problem.

Lopez, Antonio M., Jr.↗

NASA NTRS2019

Deep Learning Method for Detecting Precursors to Adverse Events

With the recent advancements in Deep Learning methods, the ability to model large complex heterogeneous data sets are fundamentally changing industry and research. Coupled with hardware improvements, and ease of implementation, a wide variety of deep neural network architectures can quickly be developed to solve a sweeping range of problems such as: object detection in images, automatic healthcare diagnosis using heterogenous data sources, real time language translating and sentence prediction, upscaling low resolution images, and forecasting of multivariate timeseries. Generally, many of these architectures outperform classical machine learning approaches in their respective tasks, however, this typically comes at a cost of interpretability. These black box algorithms generally suffer from lack of transparency in both model complexity as well as the rationale behind the prediction. This lack of comprehension, is driving an emerging area of interest in “Explainable AI”. An algorithm called: “Deep Temporal Multiple Instance Learning”1 was a recently developed to identify precursors to adverse events and has been applied in the aviation domain. The deep learning architecture is designed to capture the evolution of the probability of the outcome over the time preceding the adverse event using a multiple instance learning approach as illustrated in Figure 1. Precursors are defined when the probability of the event has exceeded a threshold at some point in the timeseries, at which point, a sensitivity analysis is performed to determine contributing factors. The contributing factors are used to explain and define the precursor during the periods where the probability score is high. The identified contributing factors are then presented to subject matter experts to provide objective insights into the leading factors associated with the particular adverse event. The algorithm has been tested on flight data from a commercial airline and has the ability to discover precursors to known adverse events that take the form of safety critical operations, such as unstable approach events on final approach. Apart from detecting precursors to adverse events, the converse can also be leveraged to discover corrective actions. These positive actions manifest themselves as periods in the timeseries when the precursor score has been lowered from an elevated state; meaning that if the system had been left uncorrected, it would have eventually reached the adverse event state. Characterizing these state changes can help identify successful interventions that may not have been known before. Policy makers and procedure designers can use this additional knowledge to craft more safety and efficient resilient procedures for future operations and therefore improve the overall performance of the National Airspace.

Matthews, Bryan L.↗

DOE OSTI2025

Architecture-Aware Models of AI Engines for High-Performance Matrix Matrix Multiplication

The AI Engine (AIE) architecture, available in systems from mobile SoCs to server-class FPGAs, aims to efficiently execute AI/ML tasks through a two-dimensional array of compute tiles. Previous work on AIEs has explored different approaches to mapping computation across spatial arrays, but the compute kernel running on each tile has not been the focus. Additionally, the AIE-ML architecture introduces memory tiles and omits programmable logic, requiring new approaches to staging and moving data throughout the array. In this work we update analytical models developed for CPUs to produce the design of high performance kernels while introducing new model considerations such as memory structure, throughput, and latency as required by the AIE hardware. We evaluate our models by developing AIE-ML kernels for matrix multiplication in low-precision data types showing performance up to 95% of compute peak for the kernel when data resides in local memory and above 90% of compute peak when data resides in main memory.

Binder, Elliott D. [Carnegie Mellon University, Pi↗

DOE OSTI2025

Autonomous Flow Electrochemistry for Accelerated Catalyst Discovery

Our objective is to develop an Autonomous Chemical Experimentation (ACE) platform that accelerates discovery of new catalytic transformations and other energy-relevant chemical reactions and processes. We intentionally designed ACE to be highly modular, both with respect to its rapid deployment to different chemistries and experimental workflows as well as incorporation of a wide range of different AI algorithms. In addition to the development of the core software architecture, initial efforts were made to incorporate Large Language Models to provide human-interpretable reasoning of the optimizer’s actions, and to develop a user-friendly graphical interface for experimental researchers. ACE was demonstrated using a flow electrocatalysis platform containing an inline FTIR spectrometer for real-time analysis and quantification of the reaction outcome. Human-in-the-loop experiments were performed in which a human researcher conducted an experiment using electrode potentials suggested by ACE, then fed the spectral data back to ACE for decision making. After confirming the successful function of the optimizer, efforts were next directed to automation of the hardware and performed full autonomy tests using three reactions: catalytic oxidation of formate, catalytic oxidation of cyclohexanol, and oxidation of hydroquinone. These studies confirm that ACE can close the loop between reaction execution, analysis, and optimization. They also reveal that more improved product detection methods will be essential for ACE to make well-informed decisions for reactions with low conversions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2022

Inspecta Annual Technical Report

Sandia National Laboratories (SNL) is designing and developing an Artificial Intelligence (AI)-enabled smart digital assistant (SDA), Inspecta (International Nuclear Safeguards Personal Examination and Containment Tracking Assistant). The goal is to provide inspectors an in-field digital assistant that can perform tasks identified as tedious, challenging, or prone to human error. During 2021, we defined the requirements for Inspecta based on reviews of International Atomic Energy Agency (IAEA) publications and interviews with former IAEA inspectors. We then mapped the requirements to current commercial or open-source technical capabilities to provide a development path for an initial Inspecta prototype while highlighting potential research and development tasks. We selected a highimpact inspection task that could be performed by an early Inspecta prototype and are developing the initial architecture, including hardware platform. This paper describes the methodology for selecting an initial task scenario, the first set of Inspecta skills needed to assist with that task scenario and finally the design and development of Inspecta’s architecture and platform.

42 ENGINEERING↗

DOE OSTI2020

5G Enabled Energy Innovation: Advanced Wireless Networks for Science (Workshop Report)

Rapidly expanding, new telecommunications infrastructure based on 5G technologies will disrupt and transform how we design, build, operate, and optimize scientific infrastructure and the experiments and services enabled by that infrastructure, from continental-scale sensor networks to centralized scientific user facilities, from intelligent Internet of Things devices to supercomputers. Concurrently, 5G will introduce, or exacerbate, challenges related to protecting infrastructure and associated scientific data as well as to fully leveraging opportunities related to expanded infrastructure scale and complexity. The U.S. Department of Energy (DOE) Office of Science operates scientific infrastructure, supporting some of the nation’s most advanced intellectual discoveries, spanning the country and including 30 world-class user facilities from supercomputers to accelerators. Along with field experiments and remote observatories, every aspect of DOE’s scientific enterprise will be affected by 5G, which amounts to a complete renovation of the underpinnings of the nation’s information infrastructure. In this report we explore the scientific opportunities and new research challenges associated with 5G, ranging from scalability to heterogeneity to cybersecurity. The rapid commercial deployment of 5G opens the opportunity to rethink and reinvent DOE’s scientific infrastructure and experimentation, from intelligent sensor networks at unprecedented scales to a digital continuum of cyberinfrastructure spanning low-power sensors, high-performance computing embedded within and at the edge of the network, and DOE’s large-scale user instrument and computing facilities. New programming paradigms, workflow and data frameworks, and AI-based system design, operation, and autonomous adaptation and optimization will be necessary in order to exploit these new opportunities. Field deployments and centralized scientific instruments can also be revolutionized, moving (without traditional performance penalties) from wired to wireless connectivity for data and control systems, improving flexibility, and opening new sensing modalities, including the use of the 5G electromagnetic spectrum itself as an environmental probe. For DOE science, in contrast to commercial 5G applications and settings, devices will be deployed in extreme environments such as cryogenically cooled instrument control systems and in remote settings with harsh conditions, requiring the design of new materials for RF communication and edge processing to operate in these regimes. Concurrently, 5G infrastructure comprises both hardware and sophisticated software systems - currently closed and proprietary. The cybersecurity challenges to 5G-empowered reinvention mirror the complexity and variety of new 5G features, from virtualization to private network slices to ubiquitous access. Research is also needed in order to accelerate the development of secure and open 5G software infrastructure, reducing reliance on hardware and software produced outside the United States and providing the transparency and rigorous evaluation and testing afforded through open software. Twelve broad research thrusts are laid out in four chapters, with a companion fifth chapter (and three additional research thrusts) underscoring the needs and opportunities for an aggressive testbed program co-designed by networking experts and scientists involved in the 15 research thrusts. The urgency of undertaking this research is fueled by a global, accelerating deployment of new telecommunications infrastructure that is designed for entertainment and commercial applications - barely scratching the surface of what 5G can do to extend U.S. leadership in scientific discovery.

42 ENGINEERING↗

NASA NTRS1988

Artificial intelligent decision support for low-cost launch vehicle integrated mission operations

The feasibility, benefits, and risks associated with Artificial Intelligence (AI) Expert Systems applied to low cost space expendable launch vehicle systems are reviewed. This study is in support of the joint USAF/NASA effort to define the next generation of a heavy-lift Advanced Launch System (ALS) which will provide economical and routine access to space. The significant technical goals of the ALS program include: a 10 fold reduction in cost per pound to orbit, launch processing in under 3 weeks, and higher reliability and safety standards than current expendables. Knowledge-based system techniques are being explored for the purpose of automating decision support processes in onboard and ground systems for pre-launch checkout and in-flight operations. Issues such as: satisfying real-time requirements, providing safety validation, hardware and Data Base Management System (DBMS) interfacing, system synergistic effects, human interfaces, and ease of maintainability, have an effect on the viability of expert systems as a useful tool.

Szatkowski, Gerard P.↗

NASA NTRS1986

Conceptual design for a food production, water and waste processing, and gas regeneration module

During the first six month period, the RECON (Regenerative Concepts Group) team collected reference material, made visits to consult with other researchers, and invited distinguished visitors to speak on the status of closed life support activities. A decision was made to develop the data base and modeling such that artificial intelligence (AI) methods could be used to manipulate data and examine concept alternatives. Six discrete tasks and a project schedule were outlined for the first year. The first two tasks have been essentially completed and have resulted in a sample set of assumptions for general use in defining candidate systems and for the specification of closed system characteristics. To model a closed environment, decisions were necessary to establish the amounts of food, air, water and waste products. Although recognized that data would eventually be normalized on the basis of a single human, the amount of data in existence for four person crews led to the decision to use this as a baseline. Information on existing concepts was collected from NASA sources, from industry, and libraries. Concept modeling was begun, hardware and software obtained, technical tasks identified and experimental work initiated.

Nicks, O. W.↗

DOE OSTI2025

Extending High-Level Synthesis with AI/ML Methods

Artificial Intelligence (AI) and Machine Learning (ML) methods provide significant opportunities of improving quality of results when performing high-level synthesis (HLS). For example, they can be used to model and predict metrics of the final design (e.g., area, considering aspects such as interconnect overhead for different device technologies), facilitating exploration when searching for the best design trade-offs. They can also enable identifying hidden correlations across the various phases of the synthesis and the various optimizations performed, identifying the most effective pipelines. Finally, in more general terms, bio-inspired heuristic algorithms can improve the design space exploration for the synthesis process in terms of time and quality of the result. This paper discusses opportunities and challenges to augment HLS with AI/ML using as example flow the SODA Synthesizer, an open-source hardware generation toolchain which includes SODA-OPT, a hardware/software partitioning and pre-optimization tool developed with the MLIR framework, and PandA-Bambu, a state-of-the art HLS tool. SODA interfaces with OpenROAD to provide a complete end-to-end toolchain.

artificial intelligence↗

DOE OSTI2025

Light-powered end-to-end neutron detection and imaging with an edge-deployed optical AI chip

Neutron detection is widely used in many applications including nuclear physics, nuclear energy, nuclear technologies and nuclear safeguards. Developing an end-to-end neutron detection and imaging workflow paves way towards fully automated processes for many applications. We implemented an automated workflow for neutron detection experiments which use a solid state image sensor to capture neutron hits as a digital image. We deploy the workflow to an edge-based optical neural network (ONN) to increase the radiation-hardness and lifetime of neutron detection instruments. We present a two-stage neural network framework for detection of neutrons at sub-pixel resolution. The first stage uses a region proposal network to efficiently detect and extract neutron hits from the input camera image. The second stage feeds the extracted hits into a fully connected neural network to predict the sub-pixel hit position. The performance of the two-stage framework is evaluated using the edge-based ONN. The results show that we can achieve above 96% neutron detection accuracy as well as sub-pixel and sub-micron position resolution, while enjoying the advantages of the ONN hardware including radiation-hardness, low energy consumption and high computing speed for integrated edge camera and hardware deployment, when compared with electronic counterparts.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2025

Using a Large Language Model as a Building Block to Generate Usable Validation and Verification Suite for OpenMP

In the HPC area, both hardware and software move quickly. Often new hardware is developed and deployed, the corresponding software stack, including compilers and other tools, are under active development while leading edge software developers are working to port and tune their applications, all at the same time. While the software ecosystem is in flux, one of the key challenges for users is obtaining insight into the state of implementation of key features in the programming languages and models their applications are using – whether they have been implemented, and whether the implementation conforms to the specification, especially for newly implemented features (less tested by widespread use). OpenMP is one of the most prominent shared memory programming models used for on-node programming in HPC. With the shift towards accelerators (such as GPUs and FPGAs) and heterogeneous programming OpenMP features are getting more complex. It is natural to ask whether generative AI approaches, and large language models (LLMs) in particular, can help in producing validation and verification test suites to allow users better and faster insights into the availability and correctness of OpenMP features of interest. In this work, we explore the use of ChatGPT-4 to generate a suite of tests for OpenMP features. We have chosen a set of directives and clauses, a total of 78 combinations, which first appeared in OpenMP 3.0 (released in May 2008) but are also relevant for accelerators. We prompted ChatGPT to generate tests in the C and Fortran languages, for both host (CPU) and device (accelerator). On the Summit super-computer using the GNU implementation, we found that, of the 78 generated tests 67 C tests and 43 Fortran tests compiled successfully and fewer than those executed to completion. On further analysis we show that not all generated tests are valid. We document the process, results, and provide detailed analysis regarding the quality of tests generated. With the aim of providing input to a production quality validation and verification suite, we manually implement the corrections required to make the tests valid according to the current OpenMP specification. We quantify this effort as small, medium, or large, and record the lines of code changed to correct the invalid tests. With the corrected tests we validate recent implementations from HPE, AMD, and GNU on the Frontier supercomputer. Our experiment and subsequent analysis show that although LLMs are capable of producing HPC specific codes, they are limited by their understanding of the deeper semantics and restrictions of programming models such as OpenMP. Unsurprisingly more commonly used features have better support, while some OpenMP 3.0 directives such as sections and tasking are not universally supported on accelerators. We demonstrate that successful compilation and execution to completion are inadequate metrics for evaluating generated code and that, at this time, commodity LLMs require expert intervention for code verification. This points to gaps in the training data that is currently available for HPC. We demonstrate that with "small" effort 37% of generated invalid C tests and 63% of generated invalid Fortran tests could be corrected. This improves productivity of test generation as we circumvent writing from scratch and the common programming errors associated with it.

Pophale, Swaroop [ORNL] (ORCID:0000000185446367)↗