Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “compute in memory”

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 415 records · Page 23
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Measuring the Energy Consumption and Efficiency of Deep Neural Networks: An Empirical Analysis and Design Recommendations

Addressing the "Red-AI" trend of rising energy consumption by large-scale neural networks, this study investigates the measured energy consumption of training various fully connected neural network architectures. We introduce the BUTTER-E dataset, an augmentation to the BUTTER Empirical Deep Learning dataset, containing energy consumption and performance data from 41,129 individual experimental runs spanning 30,582 distinct configurations: 13 datasets, 20 sizes (trainable parameters), 8 "shapes", and 14 depths on both CPUs and GPUs using node-level watt-meters. This dataset reveals the complex relationship between dataset size, network structure, and energy use. Our analysis uncovers a surprising, hardware-mediated non-linear relationship between energy efficiency and network design, challenging the assumption that reducing the number of parameters or FLOPs is the best way to achieve greater energy efficiency. We propose a straightforward and effective energy model that accounts for network size, computing, and memory hierarchy. Highlighting the need for cache-considerate algorithm development, we suggest a codesign approach to energy efficient network, algorithm, and hardware design. This work contributes to the fields of sustainable computing and Green AI, offering practical guidance for creating more energy-efficient neural networks and promoting sustainable AI.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1977

Effect of virtual memory on efficient solution of two model problems

Computers with virtual memory architecture allow programs to be written as if they were small enough to be contained in memory. Two types of problems are investigated to show that this luxury can lead to quite an inefficient performance if the programmer does not interact strongly with the characteristics of the operating system when developing the program. The two problems considered are the simultaneous solutions of a large linear system of equations by Gaussian elimination and a model three-dimensional finite-difference problem. The Control Data STAR-100 computer runs are made to demonstrate the inefficiencies of programming the problems in the manner one would naturally do if the problems were indeed, small enough to be contained in memory. Program redesigns are presented which achieve large improvements in performance through changes in the computational procedure and the data base arrangement.

Lambiotte, J. J., Jr.↗
DOE OSTI2023

A survey of techniques for optimizing transformer inference

Recent years have seen a phenomenal rise in the performance and applications of transformer neural networks. The family of transformer networks, including Bidirectional Encoder Representations from Transformer (BERT), Generative Pretrained Transformer (GPT) and Vision Transformer (ViT), have shown their effectiveness across Natural Language Processing (NLP) and Computer Vision (CV) domains. Transformer-based networks such as ChatGPT have impacted the lives of common men. However, the quest for high predictive performance has led to an exponential increase in transformers' memory and compute footprint. Researchers have proposed techniques to optimize transformer inference at all levels of abstraction. Further, this paper presents a comprehensive survey of techniques for optimizing the inference phase of transformer networks. We survey techniques such as knowledge distillation, pruning, quantization, neural architecture search and lightweight network design at the algorithmic level. We further review hardware-level optimization techniques and the design of novel hardware accelerators for transformers. We summarize the quantitative results on the number of parameters/FLOPs and the accuracy of several models/techniques to showcase the tradeoff exercised by them. We also outline future directions in this rapidly evolving field of research. We believe that this survey will educate both novice and seasoned researchers and also spark a plethora of research efforts in this field.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

From prediction to experimental realization of ferroelectric wurtzite Al1− x Gd x N alloys

AlN-based alloys find widespread application in high-power microelectronics, optoelectronics, and electromechanics. The realization of ferroelectricity in wurtzite AlN-based heterostructural alloys has opened up the possibility of directly integrating ferroelectrics with conventional microelectronics based on tetrahedral semiconductors, such as Si, SiC, and III–Vs, enabling compute-in-memory architectures, high-density data storage, and more. The discovery of AlN-based wurtzite ferroelectrics has been driven to date by chemical intuition and empirical explorations. Here, we demonstrate the computationally-guided discovery and experimental demonstration of new ferroelectric wurtzite Al1−xGdxN alloys. First-principles calculations indicate that the minimum energy pathway for switching changes from a collective to an individual switching process with a lower overall energy barrier, at a rare-earth fraction x with x > 0.10–0.15. Experimentally, ferroelectric switching is observed at room temperature in Al1−xGdxN films with x > 0.12, which strongly supports the switching mechanisms in wurtzite ferroelectrics proposed previously [Lee et al., Sci. Adv. 10, eadl0848 (2024)]. This is also the first demonstration of ferroelectricity in an AlN-based alloy with a magnetic rare-earth element, which could pave the way for additional functionalities such as multiferroicity and opto-ferroelectricity in this exciting class of AlN-based materials.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Efficient Data Management in Neutron Scattering Data Reduction Workflows at ORNL

Oak Ridge National Laboratory (ORNL) experimental neutron science facilities produce 1.2 TB a day of raw event-based data that is stored using the standard metadata-rich NeXus schema built on top of the HDF5 file format. Performance of several data reduction workflows is largely determined by the amount of time spent on the loading and processing algorithms in Mantid, an open-source data analysis framework used across several neutron sciences facilities around the world. The present work introduces new data management algorithms to address identified input output (I/O) bottlenecks on Mantid. First, we introduce an in-memory binary-tree metadata index that resemble NeXus data access patterns to provide a scalable search and extraction mechanism. Second, data encapsulation in Mantid algorithms is optimally redesigned to reduce the total compute and memory runtime footprint associated with metadata I/O reconstruction tasks. Results from this work show speed ups in wall-clock time on ORNL data reduction workflows, ranging from 11% to 30% depending on the complexity of the targeted instrument-specific data. Nevertheless, we highlight the need for more research to address reduction challenges as experimental data volumes increase.

Godoy, William↗
DOE OSTI2025

Communication Lower Bounds and Optimal Algorithms for Symmetric Matrix Computations

In this article, we focus on the communication costs of three symmetric matrix computations: (i) multiplying a matrix with its transpose, known as a symmetric rank-k update (SYRK) (ii) adding the result of the multiplication of a matrix with the transpose of another matrix and the transpose of that result, known as a symmetric rank-2k update (SYR2K) (iii) performing matrix multiplication with a symmetric input matrix (SYMM). All three computations appear in the Level 3 Basic Linear Algebra Subroutines (BLAS) and have wide use in applications involving symmetric matrices. We establish communication lower bounds for these kernels using sequential and distributed-memory parallel computational models, and we show that our bounds are tight by presenting communication-optimal algorithms for each setting. Our lower bound proofs rely on applying a geometric inequality for symmetric computations and analytically solving constrained nonlinear optimization problems. As a result, the symmetric matrix and its corresponding computations are accessed and performed according to a triangular block partitioning scheme in the optimal algorithms.

Al Daas, Hussam [Rutherford Appleton Laboratory, D↗
NASA NTRS1991

Self-Testing Static Random-Access Memory

Proposed static random-access memory for computer features improved error-detecting and -correcting capabilities. New self-testing scheme provides for detection and correction of errors at any time during normal operation - even while data being written into memory. Faults in equipment causing errors in output data detected by repeatedly testing every memory cell to determine whether it can still store both "one" and "zero", without destroying data stored in memory.

Chau, Savio↗
DOE OSTI2020

Scalable Knowledge Graph Analytics at 136 Petaflop/s

We are motivated by newly proposed methods for data mining large-scale corpora of scholarly publications, such as the full biomedical literature, which may consist of tens of millions of papers spanning decades of research. In this setting, analysts seek to discover how concepts relate to one another. They construct graph representations from annotated text databases and then formulate the relationship-mining problem as one of computing all-pairs shortest paths (APSP), which becomes a significant bottleneck. In this context, we present a new high-performance algorithm and implementation of the Floyd-Warshall algorithm for distributed-memory parallel computers accelerated by GPUs, which we call DSNAPSHOT (Distributed Accelerated Semiring All-Pairs Shortest Path). For our largest experiments, we ran DSNAPSHOT on a connected input graph with millions of vertices using 4, 096nodes (24,576GPUs) of the Oak Ridge National Laboratory's Summit supercomputer system. We find DSNAPSHOT achieves a sustained performance of 136×1015 floating-point operations per second (136petaflop/s) at a parallel efficiency of 90% under weak scaling and, in absolute speed, 70% of the best possible performance given our computation (in the single-precision tropical semiring or “min-plus” algebra). Looking forward, we believe this novel capability will enable the mining of scholarly knowledge corpora when embedded and integrated into artificial intelligence-driven natural language processing workflows at scale.

Kannan, Ramakrishnan {ramki}↗
DOE OSTI2021

A Taxonomy for Classification and Comparison of Dataflows for GNN Accelerators

Recently, Graph Neural Networks (GNNs) have received a lot of interest because of their success in learning representations from graph structured data. However, GNNs exhibit different compute and memory characteristics compared to traditional Deep Neural Networks (DNNs). Graph convolutions require feature aggregations from neighboring nodes (known as the aggregation phase), which leads to highly irregular data accesses. GNNs also have a very regular compute phase that can be broken down to matrix multiplications (known as the combination phase). All recently proposed GNN accelerators utilize different dataflows and microarchitecture optimizations for these two phases. Different communication strategies between the two phases have been also used. However, as more custom GNN accelerators are proposed, the harder it is to qualitatively classify them and quantitatively contrast them. In this work, we present a taxonomy to describe several diverse dataflows for running GNN inference on accelerators. This provides a structured way to describe and compare the design-space of GNN accelerators.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1987

Practical implementation of an accurate method for multilevel design sensitivity analysis

Solution techniques for handling large scale engineering optimization problems are reviewed. Potentials for practical applications as well as their limited capabilities are discussed. A new solution algorithm for design sensitivity is proposed. The algorithm is based upon the multilevel substructuring concept to be coupled with the adjoint method of sensitivity analysis. There are no approximations involved in the present algorithm except the usual approximations introduced due to the discretization of the finite element model. Results from the six- and thirty-bar planar truss problems show that the proposed multilevel scheme for sensitivity analysis is more effective (in terms of computer incore memory and the total CPU time) than a conventional (one level) scheme even on small problems. The new algorithm is expected to perform better for larger problems and its applications on the new generation of computer hardwares with 'parallel processing' capability is very promising.

Nguyen, Duc T.↗
DOE OSTI2025

Exploring transport-enabled gates with integrated optical addressing to demonstrate high fidelity control of trapped ion qubits in a scalable quantum computer

In recent years, experiments involving micofabricated surface ion traps have grown in complexity, and as this complexity grows, a common design has emerged in the form of quantum charge-coupled device architecture. This architecture, already utilized on multiple systems, supports multiple zones on a device for areas such as memory or computation. The shutting process between these zones is a process often seen to be minimized. An additional component to scalable surface trapped-ion experiments includes some form of integrated photonics, as free space lasers have difficulty scaling to many ions. Here, we discuss recent results in utilizing the shuttling process with integrated photonics to perform a specific type of gate, as well as demonstrating state preparation. Furthermore, we show that these gates can be utilized as a type of optical modulator as the Doppler shift that occurs during shuttling will make the light seen by shuttled ions different than that seen by stationary ions. Furthermore, we show that the shuttling operation can be utilized as an optical modulator, where the Doppler shift changes the frequency from that of a stationary ion.

42 ENGINEERING↗
DOE OSTI2022

Electron Doping-Induced Metal–Insulator Transition in LaNiO 3 and Memory Devices

LaNiO 3 (LNO) is the only rare-earth nickelate that does not undergo a temperature-driven insulator-metal transition. Here, we report the successful deposition of epitaxial LNO thin films on LaAlO 3 substrates by a chemical solution deposition method. We report a metal-insulator transition driven by electron doping from hydrogen using catalytic Pd electrodes. A series of in-plane devices with LNO thin films and Pd-Au as the two working terminals were designed and fabricated by photolithography. Further, we report their resistance switching behavior and synaptic multiresistance states under voltage pulses. The results demonstrate the universality of electron doping-driven phase transitions in the rare-earth nickelates and their relevance to emerging memory and computing technologies.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS1991

Address tracing for parallel machines

Recently implemented parallel system address-tracing methods based on several metrics are surveyed. The issues specific to collection of traces for both shared and distributed memory parallel computers are highlighted. Five general categories of address-trace collection methods are examined: hardware-captured, interrupt-based, simulation-based, altered microcode-based, and instrumented program-based traces. The problems unique to shared memory and distributed memory multiprocessors are examined separately.

Stunkel, Craig B.↗
NASA NTRS2000

Highly Parallel Computing Architectures by using Arrays of Quantum-dot Cellular Automata (QCA): Opportunities, Challenges, and Recent Results

There has been significant improvement in the performance of VLSI devices, in terms of size, power consumption, and speed, in recent years and this trend may also continue for some near future. However, it is a well known fact that there are major obstacles, i.e., physical limitation of feature size reduction and ever increasing cost of foundry, that would prevent the long term continuation of this trend. This has motivated the exploration of some fundamentally new technologies that are not dependent on the conventional feature size approach. Such technologies are expected to enable scaling to continue to the ultimate level, i.e., molecular and atomistic size. Quantum computing, quantum dot-based computing, DNA based computing, biologically inspired computing, etc., are examples of such new technologies. In particular, quantum-dots based computing by using Quantum-dot Cellular Automata (QCA) has recently been intensely investigated as a promising new technology capable of offering significant improvement over conventional VLSI in terms of reduction of feature size (and hence increase in integration level), reduction of power consumption, and increase of switching speed. Quantum dot-based computing and memory in general and QCA specifically, are intriguing to NASA due to their high packing density (10(exp 11) - 10(exp 12) per square cm ) and low power consumption (no transfer of current) and potentially higher radiation tolerant. Under Revolutionary Computing Technology (RTC) Program at the NASA/JPL Center for Integrated Space Microelectronics (CISM), we have been investigating the potential applications of QCA for the space program. To this end, exploiting the intrinsic features of QCA, we have designed novel QCA-based circuits for co-planner (i.e., single layer) and compact implementation of a class of data permutation matrices, a class of interconnection networks, and a bit-serial processor. Building upon these circuits, we have developed novel algorithms and QCA-based architectures for highly parallel and systolic computation of signal/image processing applications, such as FFT and Wavelet and Wlash-Hadamard Transforms.

Fijany, Amir↗
NASA NTRS1999

Aerodynamic Shape Optimization Using A Combined Distributed/Shared Memory Paradigm

Current parallel computational approaches involve distributed and shared memory paradigms. In the distributed memory paradigm, each processor has its own independent memory. Message passing typically uses a function library such as MPI or PVM. In the shared memory paradigm, such as that used on the SGI Origin 2000 machine, compiler directives are used to instruct the compiler to schedule multiple threads to perform calculations. In this paradigm, it must be assured that processors (threads) do not simultaneously access regions of memory in such away that errors would occur. This paper utilizes the latest version of the SGI MPI function library to combine the two parallelization paradigms to perform aerodynamic shape optimization of a generic wing/body.

Cheung, Samson↗
DOE OSTI2020

Observation of single-defect memristor in an MoS2 atomic sheet

Non-volatile resistive switching, also known as memristor effect, where an electric field switches the resistance states of a two-terminal device, has emerged as an important concept in the development of high-density information storage, computing and reconfigurable systems. The past decade has witnessed substantial advances in non-volatile resistive switching materials such as metal oxides and solid electrolytes. It was long believed that leakage currents would prevent the observation of this phenomenon for nanometre-thin insulating layers. However, the recent discovery of non-volatile resistive switching in two-dimensional monolayers of transition metal dichalcogenide and hexagonal boron nitride sandwich structures (also known as atomristors) has refuted this belief and added a new materials dimension owing to the benefits of size scaling. Here we elucidate the origin of the switching mechanism in atomic sheets using monolayer MoS 2 as a model system. Atomistic imaging and spectroscopy reveal that metal substitution into a sulfur vacancy results in a non-volatile change in the resistance, which is corroborated by computational studies of defect structures and electronic states. These findings provide an atomistic understanding of non-volatile switching and open a new direction in precision defect engineering, down to a single defect, towards achieving the smallest memristor for applications in ultra-dense memory, neuromorphic computing and radio-frequency communication systems.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2020

GentenMPI: Distributed Memory Sparse Tensor Decomposition

GentenMPl is a toolkit of sparse canonical polyadic (CP) tensor decomposition algorithms that is designed to run effectively on distributed-memory high-performance computers. Its use of distributed-memory parallelism enables it to efficiently decompose tensors that are too large for a single compute node's memory. GentenMPl leverages Sandia's decades-long investment in the Trilinos solver framework for much of its parallel-computation capability. Trilinos contains numerical algorithms and linear algebra classes that have been optimized for parallel simulation of complex physical phenomena. This work applies these tools to the data science problem of sparse tensor decomposition. In this report, we describe the use of Trilinos in GentenMPl, extensions needed for sparse tensor decomposition, and implementations of the CP-ALS (CP via alternating least squares) and GCP-SGD (generalized CP via stochastic gradient descent) sparse tensor decomposition algorithms. We show that GentenMPl can decompose sparse tensors of extreme size, e.g., a 12.6-terabyte tensor on 8192 computer cores. We demonstrate that the Trilinos backbone provides good strong and weak scaling of the tensor decomposition algorithms.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

A tensor train-based isogeometric solver for large-scale 3D poisson problems

We introduce a three-dimensional (3D), fully tensor train (TT) assembled isogeometric analysis (IGA) framework, TT-IGA, for solving partial differential equations (PDEs). Our method reformulates IGA discrete operators into TT format, enabling efficient compression and computation. Geometry evaluations use the original NURBS description at sampling points and TT approximation is applied to geometry-derived coefficient fields and discrete operators. We demonstrate the effectiveness of the proposed TT-IGA framework on the three-dimensional Poisson equation, achieving substantial reductions in memory and computational cost without compromising solution quality.

97 MATHEMATICS AND COMPUTING↗