Search NASA⌕ Search

SEARCH · Search NASA

Results for “scalable architecture”

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 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Carbon-Based Quantum Information Science with Symmetry Protected Topological States (Final Report, DOE-BES award DE-SC0023105)

This research program established the scientific foundation for the rational, bottom-up design, synthesis, isolation, and investigation of symmetry-protected topological (SPT) electron spin qubits embedded in graphene nanoribbons (GNRs). The work focused on integrating atomically precise low-dimensional carbon nanostructures with emerging quantum logic architectures, providing a pathway toward scalable quantum materials for next-generation computing and sensing technologies. A central component of the program was the elucidation of fundamental relationships between real-space molecular architecture, local spin density distributions, electronic band dispersion, and energy level alignment in atomically precise GNR systems. These correlations define key operational parameters of SPT qubits and were systematically investigated to establish quantitative benchmarks against established molecular and solid-state spin qubit platforms. Attention was given to properties critical for quantum device performance, e.g. decoherence times, spectral sharpness of energy transitions, and tunable exchange interactions between spin states. The research demonstrated that these parameters can be engineered with atomic precision through scalable bottom-up synthetic strategies. Theory-guided design played a central role in identifying candidate structures hosting topologically protected spin states. Experimental validation was performed using both ensemble measurements and single-molecule characterization. In addition to advances in quantum materials synthesis, the program developed and applied spin-sensitive scanning probe microscopy techniques capable of directly probing quantum states and dynamic processes with atomic-scale spatial resolution. These capabilities enabled direct observation and characterization of quantum structures at the single-atom level. While the research activities were primarily hypothesis-driven fundamental investigations, the program adopted a comprehensive materials-by-design framework aimed at translating scientific discoveries into technological concepts compatible with scalable and intelligent manufacturing approaches.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Hybrid Quantum–Classical Graph Transformers for Efficient Sentiment Analysis

Quantum Machine Learning (QML) offers a promising paradigm that leverages quantum computing principles to develop efficient and expressive models for learning from complex and structured data. Recent advances in natural language processing (NLP) and artificial intelligence (AI) have demonstrated capabilities in understanding, generating, and reasoning over linguistic and multimodal information. In this work, we present the Quantum Graph Transformer (QGT), a hybrid quantum–classical architecture that extends graph transformer capabilities through quantum self-attention. The QGT models variable-length sentences as token graphs, where both the embedding encoding and the self-attention mechanisms are implemented using parameterized quantum circuits (PQCs), enabling efficient contextual learning with significantly fewer trainable parameters. We train QGT using both fully connected and 𝑘 -nearest-neighbor graph structures and evaluate it on five benchmark sentiment-classification datasets. Experimental results show that QGT consistently achieves higher or comparable accuracy to existing quantum NLP models and outperforms a Classical Graph Transformer (CGT) baseline with identical architecture, achieving 29.4 × fewer parameters while requiring 3–5 × fewer samples to reach comparable performance. These findings highlight the potential of graph-based quantum models as scalable and data-efficient architectures for natural language understanding.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2021

Sensor enabled data-driven predictive analytics for modeling and control with high penetration of DERs in distribution systems

The electric power grid is undergoing a tremendous transformation due to the increasing penetration of renewable energy resources beginning with wind and more recently with the distributed energy resources (DERs) such as solar and battery storage. DERs have dramatically changed the role of the distribution systems in the overall power grid, and they are expected to contribute a significant portion of power generation in the future. If current trends for DERs continue, system operation and control will need to change dramatically for improved grid reliability and resiliency. As renewable resources increase in penetration, new and challenging operational, planning, and design problems are expected to emerge. Some of the key challenges that arise in the planning and operation of the future grid are: 1) Quantifying the impact of high DER penetration in distribution systems on bulk grid behavior over multiple time scales. 2) Identifying whether a particular DER configuration/settings have a large impact on the overall grid behavior. These challenges can be addressed in an offline manner using detailed T&D grid models and they can also be addressed in an online manner using sensor measurements. In particular, the advancement and planned growth in sensor technology in power grid over various voltage levels provide us with a unique opportunity to tackle these challenges from a data analytic perspective without needing detailed T&D grid models. A few questions that naturally arise when addressing the challenges from DERs using sensor data are: 1) How can we use limited sensor measurements to monitor & control voltage stability and small signal stability of the bulk system? 2) How can we ensure that the developed data analytic methods are robust to data availability and quality issues? 3) How can we compute the developed analytics in a scalable manner using streaming measurements? In this project, we addressed the aforementioned challenges arising from DERs and answered the questions raised above on how to effectively use the sensor measurements to enhance the reliability and performance of the electric grid. Thus, the overarching goal of this project is to develop effective reduced/representative system models from data that make the computational complexity sufficiently manageable so as to be useful to simulate, analyze, and even control complex non-linear power systems dynamics with large penetrations of DERs. In order to achieve the objective, the project team established a four-fold technical approach 1) Formulated a combined transmission-distribution co-simulation framework for data generation and validation, 2) Derived reduced/representative models of power systems based on data-driven methods for efficient computation and appropriate representation of system behavior, 3) Developed data driven characterization of power system behavior based on transfer operator theory, machine learning and optimization for model estimation, 4) Incorporated a scalable data management and processing architecture using distributed Kafka streaming applications that coordinate input data streams to the developed data analytics. The key accomplishments of the project are: 1) Development of a scalable multi-timescale T&D co-simulation framework (both for steady state and for dynamic co-simulation) using commercial solvers (PSSE and GridLAB-D). The steady-state T&D co-simulation interface is shared with our industry partner (PJM). 2) A structured reduced order dynamic model of distribution systems that can represent partial motor stalling along with a systematic procedure to derive the model parameters. 3) A PMU based online method to monitor, localize and mitigate fault-induced delayed voltage recovery using DER reactive support and load control in distribution systems. 4) Development of linear operator based robust methodologies for dynamic state estimation, uncertainty quantification, system identification and trajectory prediction for power system dynamics. 5) An adaptive damping control for utilizing wind energy resources to provide oscillation damping and system stability. 6) Implementation of Kafka-based framework for efficient processing of streaming data using Linux-based local virtual environment.

DER integration↗
DOE OSTI2020

(Invited) All-Solid-State Batteries Using Li 7 La 3 Zr 2 O 12 Garnet Electrolyte Framework

All-solid-state batteries (ASSB) using LLZO (variants of Li 7 La 3 Zr 2 O 12 ) garnet solid electrolyte with lithium metal anode potentially offer higher energy density and improved safety. Rational design of cell architecture as well as manufacturing scalability are key aspects to consider as the technology readiness level advances. A thick composite cathode layer attached to a thin solid electrolyte layer is desirable to obtain superior energy densities. Furthermore, the architecture within the composite cathode may be engineered to contain directional conduction paths of lithium ions or electrons for enhanced rate capabilities. Here, we demonstrate a functioning bulk-type LLZO based all-solid-state battery with a practical form factor incorporating the above described design concepts. Freeze-tape-casting (FTC), a scalable and environmentally friendly ceramic processing method, is used to construct 3D porous LLZO scaffolds composed of vertical arrays of LLZO walls. The thin solid-electrolyte layer is fabricated by tape-casting (TC). By sintering the stacks of FTC and TC green tapes, porous/dense bilayers and porous/dense/porous trilayers of LLZO frameworks are obtained. An ASSB was constructed using a porous/dense bilayer by infiltrating LiNi 0.6 Mn 0.2 Co 0.2 O 2 cathode powder and carbon black into the porous layer and adhering lithium metal foil to the dense side. We find it crucial to introduce a plastic crystal based soft solid electrolyte to the porous layer to electrochemically connect all cathode components, obviating the need for co-sintering to establish contact. The soft nature of the plastic crystal based solid-electrolyte may be able to accommodate the volume change of the cathode material as it cycles. In this device, which contained no liquid component, the initial discharge capacities were similar to those observed in a lithium ion battery configuration using the same cathode powder at C/10 rates.

25 ENERGY STORAGE↗
DOE OSTI2020

All Electrospray Printing of Carbon–Based Cost–Effective Perovskite Solar Cells

With the power conversion efficiencies of PSCs exceeding 25%, the perovskite solar cells (PSCs) are closer to step into the initial industrialization. Prior to transferring from laboratory fabrication to industrial manufacturing, issues such as scalability, materials cost, and production line compatibility that significantly impact the manufacturing remain to be addressed. Here we report breakthroughs on all these fronts. Carbon-based PSCs with architecture FTO/electron transport layer/perovskite/carbon, that eliminate the need for the hole transport layer and noble metal electrode, provide ultra-low-cost configuration. This PSC architecture was manufactured using a scalable and industrially compatible electrospray (ES) technique, which enables continuous printing of all the cell layers. The ES deposited electron transport layer and perovskite layer exhibited properties comparable to that of the laboratory-scale spin coating method. The ES deposited carbon electrode layer exhibited superior conductivity and interfacial microstructure in comparison to films synthesized using the conventional doctor blading technique. As a result, the fully ES printed carbon-based PSCs showed a record 14.41% power conversion efficiency, rivaling the state-of-the-art hole transporter-free PSCs. Furthermore, these results will immediately have an impact on the scalable production of PSCs.

14 SOLAR ENERGY↗
DOE OSTI2022

U-Net Decoder CRF v0.1.0

We introduce a new encoder-decoder system that overcomes adaptability and scalability issues. We adapt multiple CNNs as encoders, allowing for the definition of multiple function parameter arguments to structure the models according to the targeted datasets and scientific problem. We leverage the flexibility of the U-Net architecture to act as a scalable decoder. The CRF-RNN layer is integrated into the decoder as an optional final layer, keeping the entire system fully compatible with back-propagation.

Avaylon, Matthew↗
DOE OSTI2023

User Manual - HydraGNN: Distributed PyTorch Implementation of Multi-Headed Graph Convolutional Neural Networks

This document serves as user manual for HydraGNN, a scalable graph neural network (GNN) architecture that allows for a simultaneous prediction of multiple target properties using multi-task learning (MTL). The HydraGNN architecture is constructed by successive superposition of three different sets of layers. The first set is made of message-passing layers to exchange information across nodes in the graph and use this to update the nodal features. The second set is made of global pooling layers that aggregate information from all the nodes in the graph and map it into a scalar, and is needed only for global target properties that are related to the entire graph. The third set of layers is dedicated to the implementation of MTL, which is enabled by forking of the architecture into separate heads, each one of them dedicated to the predictive task of one specific target property. Through an object-oriented programming paradigm, HydraGNN is templated over different message-passing policies, which allows for a user-friendly hyperparameter study to assess the sensitivity of the predictive performance of the HydraGNN architecture on a specific dataset with respect to the choice of the message-passing policy. The object-oriented paradigm used by HydraGNN also allows for a user-friendly inclusion of newly developed message passing policies within the existing framework. HydraGNN supports distributed computing capabilities for scalable data reading and scalable training on leadership-class supercomputers.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Control Oriented Models for Co-Design: Technical Overview of MT HVDC, MVDC, and Solid State Transformer Building Blocks

The electric power system is shifting toward a power electronics–enabled grid, where converter based “building blocks” (e.g., high voltage direct current (HVDC) links, multi terminal HVDC (MT HVDC) networks, medium voltage DC (MVDC) links, and solid state transformers (SSTs)) provide fast, precise control of power flows, voltage, and frequency. This report develops and applies publicly shareable electromagnetic transient (EMT) and phasor models to examine how such building blocks can be composed and coordinated to support offshore wind integration, inter area transfers, feeder support, and resilience. Section 2 documents a modular multilevel converter (MMC)–based MT HVDC modeling framework and two use cases: a compact WSCC/IEEE 9 bus test system and a 240 bus “mini WECC” case with five offshore wind plants (OWFs). Phasor to EMT transfer, initialization, and sanity checks are summarized, and neutral demonstrations of normal and contingency operation are reported. Section 3 frames the problem of wind plant inertial frequency response (IFR): shaping energy release and recovery to improve nadir while avoiding aerodynamic stall; representative simulations illustrate the issues without disclosing proprietary control. Section 4 develops MVDC concepts through an IEEE 16 bus loop and an Olympic Peninsula case study that compares AC vs. MVDC corridors and shows how feeder headroom can be pooled via DC couplers. Section 5 surveys SST architectures and identifies a gap: scalable, communication free coordination of multiple SSTs for islanded feeder networks. Across the report, novel methods and configurations under separate publication and IP review are not disclosed; only topic oriented, replicable setups and non proprietary results are shown. These models and use cases are intended as foundations for future publications and co design studies on architecture, control, and coordination of PE enabled grids.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2022

Connected Loads – Grid Connected Appliances: Deployment IoT Solution for Fault Detection and Diagnostics

As one of the most energy-intensive end-uses in the commercial buildings sector, supermarkets consume around 50 kWh/ft 2 ( or 537.6 kWh/m 2 ) of electricity annually, or more than 2 million kWh of electricity per year for a typical store. The biggest consumer of energy in a supermarket is its refrigeration system, which accounts for 40–60% of its total electricity usage and is equivalent to about 2–3% of the total energy consumed by commercial buildings in United States, or around 0.5 quadrillion Btu (or 0.53 quadrillion KJ). Also, the supermarket refrigeration system is one of the biggest consumers of refrigerants. Current supermarket refrigeration systems rely on high global warming potential hydrofluorocarbon refrigerants. Reducing refrigerant usage or using environment friendly alternatives can result in significant climate benefits. Transcritical CO2 refrigeration systems have attracted more attention in recent years because of their zero-carbon emission advantages compared with traditional refrigerants. These systems are widely used in commercial buildings such as supermarkets. The refrigeration system can also be adapted to handle flexible building loads and be integrated into grid response transactive control to balance the supply and demand of the electric grid. Even minor improvements in the efficiency and operational reliability of supermarket refrigeration systems can create significant value in terms of saving energy, improving food quality, protecting the environment, reducing carbon footprint, and enhancing electric grid resilience. For decarbonization, the new administration has set a target to reduce greenhouse gas emissions by 50– 52% by 2030 and targeting a carbon-neutral economy by 2050. For electrification, the goal is to achieve 100% clean electricity by 2035. Such decarbonization and electrification in the building sector require that energy consumption in buildings be reduced significantly. Therefore, the building sector must continuously adopt new technologies to achieve its energy and carbon emission goals. One of the most fundamental technologies is the Internet of Things (IoT). IoT has proven to be an effective solution for the building domain, including building information/energy modeling, smart buildings, etc. Although much progress has been made in the development of IoT-based building energy systems, there is still a lack of reliable, scalable, and affordable IoT-based automated fault and degradation diagnostics (AFDDs) solutions. Such solutions would enable deployment of advanced algorithms into real systems to archive the projected energy benefits. This study reviews existing IoT solutions developed for building energy– related application and developed a simple but effective AFDD IoT deployment solution, including developing a suitable IoT architecture and conducting easy and scalable deployment by leveraging a common cloud-based IoT service.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2022

Automated fault detection and diagnosis deployment Internet of Things solution for building energy system

The building sector is the highest energy consumer and ranks first in terms of carbon emissions among all sectors. To address these issues, decarbonization and electrification in the building energy sector are two critical missions of the new US administration. For decarbonization, the new administration has set a target to reduce greenhouse gas emissions by 50–52% by 2030 and targeting a carbon-neutral economy by 2050. For electrification, the goal is to achieve 100% clean electricity by 2035. Such decarbonization and electrification in the building sector require that energy consumption in buildings be reduced significantly. Therefore, the building sector must continuously adopt new technologies to achieve its energy and carbon emission goals. One of the most fundamental technologies is the Internet of Things (IoT). IoT has proven to be an effective solution for the building domain, including building information/energy modeling, smart buildings, etc. Although much progress has been made in the development of IoT-based building energy systems, there is still a lack of reliable, scalable, and affordable IoT-based automated fault and degradation diagnostic (AFDD) solutions. Such solutions would enable deployment of advanced algorithms into real systems to archive the projected energy benefits. Here, this study reviews existing IoT solutions developed for building energy–related application and developed a simple but effective AFDD IoT deployment solution, including developing a suitable IoT architecture and conducting easy and scalable deployment by leveraging a common cloud-based IoT service.

42 ENGINEERING↗
DOE OSTI2022

Integration of Microgrids, DER Aggregators, and DERMS with ADMS

Distributed energy resource (DER) management and control is disjointed, siloed, and at times conflicting. Behind-the-meter (BTM) assets can provide significant flexibility but are poorly integrated with the grid. Centralized control methods alone are not scalable. The objective is to develop and demonstrate an architecture that: provides reliable, resilient, and secure grid services; enables scalable, near-real-time management of utility- and small-scale DERs; supports transactive control, aggregation, and direct control of DERs; and incorporates existing utility management systems.

ADMS↗
DOE OSTI2023

Subspace recursive Fermi-operator expansion strategies for large-scale DFT eigenvalue problems on HPC architectures

Quantum mechanical calculations for material modeling using Kohn–Sham density functional theory (DFT) involve the solution of a nonlinear eigenvalue problem for N smallest eigenvector-eigenvalue pairs, with N proportional to the number of electrons in the material system. Here, these calculations are computationally demanding and have asymptotic cubic scaling complexity with the number of electrons. Large-scale matrix eigenvalue problems arising from the discretization of the Kohn–Sham DFT equations employing a systematically convergent basis traditionally rely on iterative orthogonal projection methods, which are shown to be computationally efficient and scalable on massively parallel computing architectures. However, as the size of the material system increases, these methods are known to incur dominant computational costs through the Rayleigh–Ritz projection step of the discretized Kohn–Sham Hamiltonian matrix and the subsequent subspace diagonalization of the projected matrix. This work explores the potential of polynomial expansion approaches based on recursive Fermi-operator expansion as an alternative to the subspace diagonalization of the projected Hamiltonian matrix to reduce the computational cost. Subsequently, we perform a detailed comparison of various recursive polynomial expansion approaches to the traditional approach of explicit diagonalization on both multi-node central processing unit and graphics processing unit architectures and assess their relative performance in terms of accuracy, computational efficiency, scaling behavior, and energy efficiency.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Targeted Writing and Deleting of Magnetic Skyrmions in Two-Terminal Nanowire Devices

Controllable writing and deleting of nanoscale magnetic skyrmions are key requirements for their use as information carriers for next-generation memory and computing technologies. While several schemes have been proposed, they require complex fabrication techniques or precisely tailored electrical inputs, which limits their long-term scalability. In this work, we demonstrate an alternative approach for writing and deleting skyrmions using conventional electrical pulses within a simple, two-terminal wire geometry. X-ray microscopy experiments and micromagnetic simulations establish the observed skyrmion creation and annihilation as arising from Joule heating and Oersted field effects of the current pulses, respectively. The unique characteristics of these writing and deleting schemes, such as spatial and temporal selectivity, together with the simplicity of the two-terminal device architecture, provide a flexible and scalable route to the viable applications of skyrmions.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2025

Nano-Engineered Interfaces in Dual-Layer Electrodes for Protonic Ceramic Cells with Enhanced Stability and Kinetics

Enhancing interfacial stability and charge transfer in protonic ceramic cells (PCCs) remains a critical challenge, as structural degradation and interfacial resistance often compromise durability and efficiency. Here, we report a nanoengineered dual-layer oxygen electrode architecture designed to address these limitations by introducing a fine-grained nanoparticle interfacial contact layer beneath a porous catalytic backbone. The nanoscale powders, through enhanced sintering activity, densify into a robust interfacial layer that promotes strong chemical bonding, uniform adhesion, and continuous ionic/electronic pathways with the BCZYYb electrolyte. This hierarchical architecture mitigates delamination, redistributes mechanical stress, and establishes efficient charge and mass transport channels without relying on corrosive surface treatments. Electrochemical evaluation demonstrates that the dual-layer design markedly reduces interfacial polarization resistance and accelerates electrode kinetics. Compared to the single-layer counterpart, the architecture achieves a peel strength of 44.53 N/cm 2 , a 40% improvement in peak power density (0.96 W cm –2 at 600 °C), and a 130% enhancement in electrolysis current density (4.78 A cm –2 at 1.57 V). Faradaic efficiency remains as high as 88% under high steam concentrations, underscoring minimal charge loss during practical operation. Notably, the electrode retains stability across 450–600 °C and under transient voltage cycling, with impedance spectra confirming suppressed interfacial resistance growth over prolonged use. These results highlight nanoscale interface engineering as a powerful route to enhance both mechanical robustness and electrochemical kinetics in PCCs. The demonstrated scalability and durability of this architecture provide a versatile platform for advancing solid-state electrochemical systems, including reversible fuel cells and high-efficiency hydrogen production technologies.

Faradaic efficiency↗
DOE OSTI2026

User Manual - HydraGNN v5.0: Distributed Implementation of Multi-Tasking Graph Neural Networks

This document serves as the user manual for HydraGNN v5.0, a scalable graph neural network (GNN) architecture for simultaneous prediction of multiple target properties using multi-task learning (MTL). This version of HydraGNN has been developed primarily to support the development, training, and deployment of predictive graph-based deep learning (DL) models for atomistic materials modeling. HydraGNN is templated over 13 message-passing policies, including invariant models (GIN, PNA, PNAPlus, GAT, MFC, CGCNN, SAGE, SchNet, DimeNet) and equivariant models (EGNN, PNAEq, PAINN, MACE), and supports distributed training via distributed data parallelism (DDP), DeepSpeed, and Fully Sharded Data Parallelism (FSDP) on leadership-class supercomputers. Although HydraGNN can be applied to problems beyond atomistic materials modeling, its current use is confined to homogeneous graphs. Additional capabilities include machine-learned interatomic potentials with energy-conserving forces, General, Powerful, and Scalable Graph Transformer (GraphGPS) global attention, periodic boundary conditions, hyperparameter optimization, mixed-precision training, and uncertainty quantification.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Hierarchical Model-Free Transactional Control of Building Loads to Support Grid Services

A transition from generation on demand to consumption on demand is one of the solutions to overcome the many limitations associated with the higher penetration of renewable energy sources. Such a transition, however, requires a considerable amount of load flexibility in the demand side. Demand response (DR) programs can reveal and utilize this demand flexibility by enabling the participation of a large number of grid-interactive efficient buildings (GEB). Existing approaches on DR require significant modelling or training efforts, are computationally expensive, and do not guarantee the satisfaction of end users. To address the aforementioned limitations, this software code considers a scalable hierarchical model-free transactional control approach that incorporates elements of virtual battery, game theory, and model-free control (MFC) mechanisms. The developed approach separates the control mechanism into upper and lower levels. The MFC modulates the flexible GEB in the lower level with guaranteed thermal comfort of end users in response to the optimal pricing and power signals determined in the upper layer using a Stackelberg game integrated with aggregate virtual battery constraints. Additionally, the usage of MFC necessitates less burdensome computational and communication requirements, thus, it is easily deployable even on small, embedded devices. This software code enables the use of residential and small-size commercial buildings to offer a potentially substantial source of ancillary grid services that are currently underutilized. A hierarchical, model-free transactive building control provides a smooth interface between the grid service requests of utilities and the reliable control required by participating buildings. Model-free control, which supports distributed control architecture, will be a more scalable solution that can be deployed in neighborhood-size systems as well as individual buildings. The proposed approach contributes to the body of knowledge on two main aspects. First, it couples the MFC with the game-theoretic control and proposes a scalable model-free transactive control approach. MFC does not require any modeling effort or model training for the various building loads. This is very beneficial since deriving an accurate model for every single unit participating in DR programs and obtaining all the parameters about the units (e.g., thermal coefficients, standby losses) are infeasible. Also, MFC is computationally efficient, easily deployable even on small, embedded devices, and can be implemented in real time. Second, it integrates the concept of virtual battery into DR via the Stackelberg game. The concept of virtual battery enables efficient coordination and aggregation of a large number of flexible GEB with guaranteed thermal comfort of end users.

Olama, Mohammed [Oak Ridge National Lab. (ORNL), O↗
DOE OSTI2021

Machine Learning-enabled Scalable Performance Prediction of Scientific Codes

Hardware architectures become increasingly complex as the compute capabilities grow to exascale. Here, we present the Analytical Memory Model with Pipelines (AMMP) of the Performance Prediction Toolkit (PPT). PPT-AMMP takes high-level source code and hardware architecture parameters as input and predicts runtime of that code on the target hardware platform, which is defined in the input parameters. PPT-AMMP transforms the code to an (architecture-independent) intermediate representation, then (i) analyzes the basic block structure of the code, (ii) processes architecture-independent virtual memory access patterns that it uses to build memory reuse distance distribution models for each basic block, and (iii) runs detailed basic-block level simulations to determine hardware pipeline usage. PPT-AMMP uses machine learning and regression techniques to build the prediction models based on small instances of the input code, then integrates into a higher-order discrete-event simulation model of PPT running on Simian PDES engine. We validate PPT-AMMP on four standard computational physics benchmarks and present a use case of hardware parameter sensitivity analysis to identify bottleneck hardware resources on different code inputs. We further extend PPT-AMMP to predict the performance of a scientific application code, namely, the radiation transport mini-app SNAP. To this end, we analyze multi-variate regression models that accurately predict the reuse profiles and the basic block counts. We validate predicted SNAP runtimes against actual measured times.

97 MATHEMATICS AND COMPUTING↗