Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Computer Operations and 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.

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

At least 145 records · Page 8
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Encoding integers and rationals on neuromorphic computers using virtual neuron

Abstract Neuromorphic computers emulate the human brain while being extremely power efficient for computing tasks. In fact, they are poised to be critical for energy-efficient computing in the future. Neuromorphic computers are primarily used in spiking neural network–based machine learning applications. However, they are known to be Turing-complete, and in theory can perform all general-purpose computation. One of the biggest bottlenecks in realizing general-purpose computations on neuromorphic computers today is the inability to efficiently encode data on the neuromorphic computers. To fully realize the potential of neuromorphic computers for energy-efficient general-purpose computing, efficient mechanisms must be devised for encoding numbers. Current encoding mechanisms (e.g., binning, rate-based encoding, and time-based encoding) have limited applicability and are not suited for general-purpose computation. In this paper, we present the virtual neuron abstraction as a mechanism for encoding and adding integers and rational numbers by using spiking neural network primitives. We evaluate the performance of the virtual neuron on physical and simulated neuromorphic hardware. We estimate that the virtual neuron could perform an addition operation using just 23 nJ of energy on average with a mixed-signal, memristor-based neuromorphic processor. We also demonstrate the utility of the virtual neuron by using it in some of the μ -recursive functions, which are the building blocks of general-purpose computation.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

Open Source Software for HPC

The computational power of HPC is beyond our comprehension when we hear that 5 quadrillion computations can happen in a matter of seconds, or that machine learning is changing the way everything works. But none of that happens in a vacuum, and the teams behind the scenes—the developers of the hardware, the operating systems, the data transfer protocols, and the applications themselves—are the unsung heroes of a world where faster is better and you'd better hope there's no bug in the software or the hardware to slow you down. HPC is most successful when all these aspects work together seamlessly. The stories that follow are a tribute to the hardworking teams behind the scenes.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Calibrating the Classical Hardness of the Quantum Approximate Optimization Algorithm

The trading of fidelity for scale enables approximate classical simulators such as matrix product states (MPSs) to run quantum circuits beyond exact methods. A control parameter, the so-called bond dimension $\mathcal{χ}$ for MPSs, governs the allocated computational resources and the output fidelity. Here, we characterize the fidelity for the quantum approximate optimization algorithm by the expectation value of the cost function that it seeks to minimize and find that it follows a scaling law $\mathscr{F}$(ln $\mathcal{χ}$/N), where N is the number of qubits. With ln $\mathcal{χ}$ amounting to the entanglement that a MPS can encode, we show that the relevant variable for investigating the fidelity is the entanglement per qubit. Importantly, our results calibrate the classical computational power required to achieve the desired fidelity and benchmark the performance of quantum hardware in a realistic setup. For instance, we quantify the hardness of performing better classically than a noisy superconducting quantum processor by readily matching its output to the scaling function. Moreover, we relate the global fidelity to that of individual operations and establish its relationship with $\mathcal{χ}$ and N. We sharpen the requirements for noisy quantum computers to outperform classical techniques at running a quantum optimization algorithm in speed, size, and fidelity.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2013

Power/energy use cases for high performance computing

Power and Energy have been identified as a first order challenge for future extreme scale high performance computing (HPC) systems. In practice the breakthroughs will need to be provided by the hardware vendors. But to make the best use of the solutions in an HPC environment, it will likely require periodic tuning by facility operators and software components. This document describes the actions and interactions needed to maximize power resources. It strives to cover the entire operational space in which an HPC system occupies. The descriptions are presented as formal use cases, as documented in the Unified Modeling Language Specification [1]. The document is intended to provide a common understanding to the HPC community of the necessary management and control capabilities. Assuming a common understanding can be achieved, the next step will be to develop a set of Application Programing Interfaces (APIs) to which hardware vendors and software developers could utilize to steer power consumption.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Nonunitary Variational Quantum Eigensolver with the Localized Active Space Method and Cost Mitigation

Accurately describing strongly correlated systems with affordable quantum resources remains a central challenge for quantum chemistry applications on near and intermediate term quantum computers. The localized active space self-consistent field (LASSCF) approximates the complete active space self-consistent field (CASSCF) by generating active space-based wave functions within specific fragments while treating interfragment correlation with mean-field approach, hence is computationally less expensive. Hardware-efficient ansatzes (HEA) offer affordable and shallower circuits, yet they often fail to capture the necessary correlation. Previously, Jastrow-factor-inspired nonunitary qubit operators were proposed to use with HEA for variational quantum eigensolver (VQE) calculations (so-called nuVQE), as they do not increase circuit depths and recover correlation beyond the mean-field level for Hartree–Fock initial states. Here, in this study, we explore running nuVQE with LASSCF as the initial state. The method, named LAS-nuVQE, is shown to recover interfragment correlations, reach chemical accuracy with a small number of gates (<70) in both H 4 and square cyclobutadiene (C 4 H 4 ), and produces more accurate energetics than its HEA counterparts at all circuit depths. To further address the inherent symmetry-breaking in HEA, we implemented spin-constrained LAS-nuVQE to extend the capabilities of HEA further and show spin-pure results for square cyclobutadiene. We also mitigate the increased measurement overhead of nuVQE via Pauli grouping and shot-frugal sampling, reducing measurement costs by up to 2 orders of magnitude compared to ungrouped operator, and show that one can achieve better accuracy with a small number of shots (10 3–4 ) per one expectation value calculation compared to noiseless simulations with one or two orders of magnitude more shots. Finally, wall clock time estimates show that, with our measurement mitigation protocols, nuVQE becomes a cheaper and more accurate alternative than vanilla VQE with HEA. Taken together, these developments illustrate a practical pathway toward performing multireference chemical simulations with accuracy and affordable resources on today’s quantum hardware, achieving both accuracy and affordability in challenging correlated systems.

Wang, Qiaohong [Univ. of Chicago, IL (United State↗
DOE OSTI2024

3-center and 4-center 2-particle Gaussian AO integrals on modern accelerated processors

We report an implementation of the McMurchie–Davidson (MD) algorithm for 3-center and 4-center 2-particle integrals over Gaussian atomic orbitals (AOs) with low and high angular momenta l and varying degrees of contraction for graphical processing units (GPUs). This work builds upon our recent implementation of a matrix form of the MD algorithm that is efficient for GPU evaluation of 4-center 2-particle integrals over Gaussian AOs of high angular momenta (l ≥ 4) [A. Asadchev and E. F. Valeev, J. Phys. Chem. A 127, 10889–10895 (2023)]. The use of unconventional data layouts and three variants of the MD algorithm allow for the evaluation of integrals with double precision and sustained performance between 25% and 70% of the theoretical hardware peak. Performance assessment includes integrals over AOs with l ≤ 6 (a higher l is supported). Preliminary implementation of the Hartree–Fock exchange operator is presented and assessed for computations with up to a quadruple-zeta basis and more than 20 000 AOs. The corresponding C++ code is part of the experimental open-source LibintX library available at https://github.com/ValeevGroup/libintx.

Chemistry↗
DOE OSTI2019

TestSNAP

TestSNAP is a stand-alone prototype of the SNAP force kernel used in the LAMMPS molecular dynamics code. It contains synthetic input data (neighbor atom positions) and reference output data (forces on atoms) for several different SNAP models. The prototype performs the force calculation repeatedly, checking the results against the reference data. At the end it reports the grind time [msec/atomstep] and RMS force error [eV/A]. This allows performance experts to quickly test new hardware and software strategies, obtain a realistic performance estimate and verify that the computation was performed correctly. SAND2019-12243 M Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-NA0003525.

Thompson, Aidan↗
DOE OSTI2021

Computational Fluid Dynamics Simulations to Support Efficiency Improvements in Aluminum Smelting Process

Smelting is broadly described as the extraction of a metal from its ore. In the United States, aluminum is commonly produced by smelting alumina in bauxite using the Hall-Héroult process. Optimization of equipment and processes in conventional smelting is crucial to enhancing process efficiency and productivity, is necessary for improving the techno-economic feasibility, which directly manifests as the growth of the American economy. To achieve optima, insightful data on the multiphysics phenomena that are inherent to the process must be obtained through physical investigation or high-fidelity numerical simulations. The resolution of relevant scales in time and space for smelting operations requires intensive, high-performance computing (HPC) simulations. Hostile operating conditions limit physical data acquisition to specific techniques; therefore, these data do not describe the multiscale interaction of simultaneous effects. Fortunately, in recent decades, significant advancements in computing hardware and computational methods have made the numerical resolution of such a complex process possible. In this study, a high-fidelity simulation of aluminum smelting was performed using an open-source tool, OpenFOAM, which analyzed many parameters characteristic to underlying phenomena. A multiphysics model based on the Eulerian-Eulerian multifluid approach was adopted. This model can resolve critical issues in the electrolytic smelting of aluminum, such as bubbling of carbon dioxide from the anode(s), magnetohydrodynamics from electromagnetic effects, ionic dissolution of the alumina in the electrolyte, and the evolution of thermal profiles. This study provides valuable connectivity for characteristic data that can direct the future designs of efficient smelters. A basic framework to model and simulate the smelting process using OpenFOAM is presented for user modification in keeping with process development. Of relevance to the flow field, a detailed investigation of vortices produced by bubble motion and electromagnetics is discussed, along with their impact on the evolution of thermal profiles. The predictions show small-scale vortices in the clearance between the anode and cathode caused by magnetic forces. Predictions also indicate relatively large-scale vortices in the inter-anode space resulting from carbon dioxide rising through the electrolytic flow field. The formation of vortices at the edges of anodes was shown to direct alumina charged by the feeder to the bottom of the anodes, thus preventing the entrapment of gas bubbles in the periphery of the bottom of the anode. Symmetry was observed in the location of cold spots in the electrolytic mixture in the vicinity of the feeder. Cold spots were also observed in the clearance between the anode and cathode due to the flow’s transmission of unconverted alumina to this region.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Practical Implementation of GPU-based Computing at the Grid Edge for Resilience Scenarios

This paper presents a practical implementation of GPU-accelerated computing at the grid edge to enhance power system resilience through next-generation smart meters. Advanced Metering Infrastructure (AMI) systems rely predominantly on centralized processing architectures, which limit real-time response capabilities during grid disturbances. This work proposes the integration of GPU-enabled computational platforms directly within smart meter to enable local execution support for power system analytics, fault detection algorithms, and optimization routines. The proposed framework uses the Julia programming language to leverage highperformance parallel computing capabilities while maintaining code portability and development efficiency. We use two experimental scenarios to benchmark the computational feasibility of this approach: sparse linear system solutions representative of power flow analyses, and multi-stage production cost simulations incorporating unit commitment and economic dispatch operations. Results demonstrate that computationally intensive power system algorithms, such as those supporting resilience scenario calculations, can be effectively executed at the distribution edge using commercially available embedded GPU hardware. Keywords—GPU acceleration, edge computing, smart meters, grid resilience, AMI, resilience.

De Souza, Reubun [School of Electrical Engineering↗
DOE OSTI2017

Beyond 100 Gb/s: Capacity, Flexibility, and Network Optimization

Here, we discuss building blocks that enable the exploitation of optical capacities beyond 100 Gb/s. Optical networks will benefit from more flexibility and agility in their network elements, especially from coherent transceivers. To achieve capacities of 400 Gb/s and more, coherent transceivers will operate at higher symbol rates. This will be made possible with higher bandwidth components using new electro-optic technologies implemented with indium phosphide and silicon photonics. Digital signal processing will benefit from new algorithms. Multi-dimensional modulation, of which some formats are already in existence in current flexible coherent transceivers, will provide improved tolerance to noise and fiber nonlinearities. Constellation shaping will further improve these tolerances while allowing a finer granularity in the selection of capacity. Frequency-division multiplexing will also provide improved tolerance to the nonlinear characteristics of fibers. Algorithms with reduced computation complexity will allow the implementation, at speeds, of direct pre-compensation of nonlinear propagation effects. Advancement in forward error correction will shrink the performance gap with Shannon's limit. At the network control and management level, new tools are being developed to achieve a more efficient utilization of networks. This will also allow for network virtualization, orchestration, and management. Lastly, FlexEthernet and FlexOTN will be put in place to allow network operators to optimize capacity in their optical transport networks without manual changes to the client hardware.

47 OTHER INSTRUMENTATION↗
DOE OSTI2016

On the performance and energy efficiency of sparse linear algebra on GPUs

In this paper we unveil some performance and energy efficiency frontiers for sparse computations on GPU-based supercomputers. We compare the resource efficiency of different sparse matrix–vector products (SpMV) taken from libraries such as cuSPARSE and MAGMA for GPU and Intel’s MKL for multicore CPUs, and develop a GPU sparse matrix–matrix product (SpMM) implementation that handles the simultaneous multiplication of a sparse matrix with a set of vectors in block-wise fashion. While a typical sparse computation such as the SpMV reaches only a fraction of the peak of current GPUs, we show that the SpMM succeeds in exceeding the memory-bound limitations of the SpMV. We integrate this kernel into a GPU-accelerated Locally Optimal Block Preconditioned Conjugate Gradient (LOBPCG) eigensolver. LOBPCG is chosen as a benchmark algorithm for this study as it combines an interesting mix of sparse and dense linear algebra operations that is typical for complex simulation applications, and allows for hardware-aware optimizations. In a detailed analysis we compare the performance and energy efficiency against a multi-threaded CPU counterpart. The reported performance and energy efficiency results are indicative of sparse computations on supercomputers.

Anzt, Hartwig↗
DOE OSTI2020

A phenomenological rate of injection model for predicting fuel injection with application to mixture formation in light-duty diesel engines

Fuel injection rate laws are one of the most important pieces of information needed when modeling engine combustion with computational fluid dynamics. In this study, a simple phenomenological model of a common-rail injector was developed and calibrated for the Bosch CRI2.2 platform. The model requires three tunable parameter fits, making it relatively easy to calibrate and suitable for injector modeling when high-fidelity information about the internal injector’s geometry and electrical circuit details are not available. Each injection pulse is modeled as a sequence of up to four stages: an injection needle mechanical opening transient, a full-lift viscous flow inertial transient, a Bernoulli steady-state stage, and a needle descent transient. Parameters for each stage are obtained as polynomial fits from measured injection rate properties. The model enforces total injected mass, and the intermediate stages are only introduced if the injection pulse duration is long enough. Experimental rates of injection from two separate campaigns on the same injector were used to calibrate the model. The model was first validated against measured injection rate laws featuring pilot injections, short partially premixed combustion pulses, and conventional diesel combustion injection strategies. Then, it was employed as an input to engine computational fluid dynamics simulations, which were run to simulate experiments of mixture formation in an optically accessible light-duty diesel engine. It was found that, though simple, this model is capable of predicting both pilot and main injection pulse mass flow rates well: the simulations yielded accurate predictions of in-cylinder equivalence ratio distributions from injection strategies for both partially premixed combustion and pilot injections. Also, once calibrated, the model produced appropriate results for a wide range of injected mass and rail pressure values. Finally, it was observed that usage of such a relatively simple model can be a good choice when high-fidelity injection rate input and highly detailed information of the injector’s geometry and operation are not available, particularly as noticeable discrepancies can be present also among different experimental campaigns on similar hardware.

42 ENGINEERING↗
DOE OSTI2017

General Mode Scanning Probe Microscopy

A critical part of SPM measurements is the information transfer from the probe-sample junction to the measurement system. Current information transfer methods heavily compress the information-rich data stream by averaging the data over a time interval, or via heterodyne detection approaches such as lock-in amplifiers and phase-locked loops. As a consequence, highly valuable information at the sub-microsecond time scales or information from frequencies outside the measurement band is lost. We have developed a fundamentally new approach called General Mode (G-mode), where we can capture the complete information stream from the detectors in the microscope. The availability of the complete information allows the microscope operator to analyze the data via information-theory analysis or comprehensive physical models. Furthermore, the complete data stream enables advanced data-driven filtering algorithms, multi-resolution imaging, ultrafast spectroscropic imaging, spatial mapping of multidimensional variability in material properties, etc. Though we applied this approach to scanning probe microscopy, the general philosophy of G-mode can be applied to many other modes of microscopy. G-mode data is captured by completely custom software written in LabVIEW and Matlab. The software generates the waveforms to electrically, thermally, or mechanically excite the SPM probe. It handles real-time communications with the microscope software for operations such as moving the SPM probe position and also controls other instrumentation hardware. The software also controls multiple variants of high-speed data acquisition cards to excite the SPM probe with the excitation waveform and simultaneously measure multiple channels of information from the microscope detectors at sampling rates of 1-100 MHz. The software also saves the raw data to the computer and allows the microscope operator to visualize processed or filtered data during the experiment. The software performs all these features while offering a user-friendly interface.

Somnath, Suhas↗
DOE OSTI2014

REPORT OF RESEARCH ACCOMPLISHMENTS AND FUTURE GOALS HIGH ENERGY PHYSICS

Caltech High Energy Physics (HEP) has a broad program in both experimental and theoretical physics. We are known for our creativity and leadership. The future is uncertain and we strive to be involved in all the major areas of experimental and theoretical HEP physics so no matter where the important discoveries occur we are well positioned to play an important role. An outstanding group of postdoctoral scholars, graduate students, staff scientists, and technical and administrative personnel support our efforts in experimental and theoretical physics. The PI’s on this grant are involved in the following program of experimental and theoretical activities: I) EXPERIMENTAL PHYSICS Our CMS group, led by Harvey Newman and Maria Spiropulu, has played a key role in the discovery and interpretation of the Higgs boson and in searches for new physics. They have important hardware responsibilities in both ECAL and HCAL and are also involved in the upgrades needed for the High Luminosity LHC. Newman's group also develops and operates Grid-based computing, networking, and collaborative systems for CMS and the US HEP community. The charged lepton (Mu2e) and quark BaBar flavor physics group is led by David Hitlin and Frank Porter. On Mu2e they have been instrumental in the design of the calorimeter. Construction responsibilities include one third of the crystals and associated readout as well as the calibration system. They also will have responsibility for a major part of the online system software. Although data taking ceased in 2008 the Caltech BaBar group is active on several new forefront analyses. The neutrino group is led by Ryan Patterson. They are central to NOvA's core oscillation physics program, to calibration, and to detector readiness being responsible for the production and installation of 12,000 APD arrays. They have key roles in neutrino appearance and disappearance analysis in MINOS and MINOS+. Sunil Golwala leads the dark matter direct detection effort. Areas of activity include: CDMS II data analysis, contributions to SuperCDMS Soudan operations and analysis, R&D towards SuperCDMS SNOLAB, development of a novel screener for radiocontamination (the BetaCage), and development of new WIMP detector concepts. Ren-Yuan Zhu leads the HEP crystal laboratory for the advanced detector R&D effort. The crystal lab is involved in development of novel scintillating crystals and has proposed several crystal based detector concepts for future HEP experiments at the energy and intensity frontiers. Its current research effort is concentrated on development of fast crystal scintillators with good radiation hardness and low cost. II) THEORETICAL PHYSICS The main theme of Sergei Gukov's current research is the relation between the geometry of quantum group invariants and their categorification, on the one hand, and the physics of supersymmetric gauge theory and string theory, on the other. Anton Kapustin's research spans a variety of topics in non-perturbative Quantum Field Theory (QFT). His main areas of interest are supersymmetric gauge theories, non-perturbative dualities in QFT, disorder operators, Topological Quantum Field Theory, and non-relativistic QFT. He is also interested in the foundations and possible generalizations of Quantum Mechanics. Hirosi Ooguri's current research has two main components. One is to find exact results in Calabi-Yau compactification of string theory. Another is to explore applications of the AdS/CFT correspondence. He also plans to continue his project with Caltech postdoctoral fellows on BPS spectra of supersymmetric gauge theories in diverse dimensions. John Preskill works on quantum information science. This field may lead to important future technologies, and also lead to new understanding of issues in fundamental physics John Schwarz has been exploring a number of topics in superstring theory/M-theory, supersymmetric gauge theory, and their AdS/CFT relationships. Much of the motivation for these studies is the desire to gain a deeper understanding of superstring theory and M-theory. The research interests of Mark Wise span particle physics, cosmology and nuclear physics. His recent work has centered on extensions of the standard model where baryon number and lepton number are gauged and extensions of the standard model that have novel sources of baryon number violation and new sources of charged lepton flavor violation

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2013

Implementing the Data Center Energy Productivity Metric in a High Performance Computing Data Center

As data centers proliferate in size and number, the improvement of their energy efficiency and productivity has become an economic and environmental imperative. Making these improvements requires metrics that are robust, interpretable, and practical. We discuss the properties of a number of the proposed metrics of energy efficiency and productivity. In particular, we focus on the Data Center Energy Productivity (DCeP) metric, which is the ratio of useful work produced by the data center to the energy consumed performing that work. We describe our approach for using DCeP as the principal outcome of a designed experiment using a highly instrumented, high-performance computing data center. We found that DCeP was successful in clearly distinguishing different operational states in the data center, thereby validating its utility as a metric for identifying configurations of hardware and software that would improve energy productivity. We also discuss some of the challenges and benefits associated with implementing the DCeP metric, and we examine the efficacy of the metric in making comparisons within a data center and between data centers.

Sego, Landon H.↗
DOE OSTI2014

Development of a versatile hybrid testing system for seismic experimentation

The hybrid testing method for earthquake engineering was developed to evaluate the seismic performance of civil structural systems. During hybrid testing, part of the structure, called physical substructure, is physically tested using hydraulic loading equipment while the rest of the structure, named numerical substructure, is numerically simulated with a computer model. Thus, hybrid testing allows a complex physical substructure being tested experimentally, while the relatively simple part of the structure is numerically simulated to economically obtain the full structural response. Recently, a versatile hybrid testing system was built at Western Michigan University consisting of a shake table, an actuator/reaction system, and an advanced hybrid testing controller. Such a testing system is capable of conducting various seismic experiments such as displacement-based pseudodynamic substructure testing as well as force-based real time dynamic hybrid testing. Because of its versatility and easy operation at the benchmark scale, the developed testing system is particularly suitable for the development of advanced hybrid testing techniques and, earthquake engineering education/outreach activities. The development of the testing system is described in this article focusing on the hardware and software integrations. A three story shake table hybrid test is presented as an example to demonstrate the system’s capability and feasibility.

42 ENGINEERING↗
DOE OSTI2025

Quantum computing for energy correlators

In recent years, energy correlators have emerged as powerful observables for probing the fragmentation dynamics of high-energy collisions. We introduce the first numerical strategy for calculating energy correlators using the Hamiltonian lattice approach, providing access to the intriguing nonperturbative dynamics of these observables. Furthermore, motivated by rapid advances in quantum computing hardware and algorithms, we propose a quantum algorithm for calculating energy correlators in quantum field theories. This algorithm includes ground state preparation, the application of source, sink, energy flux and real-time evolution operators, and the Hadamard test. We validate our approach by applying it to the SU(2) pure gauge theory in 2 + 1 dimensions on 3 × 3 and 5 × 5 honeycomb lattices with 𝑗 max = $\frac{1}{2}$ at various couplings, utilizing both classical methods and the quantum algorithm, the latter tested using the IBM emulator for specific configurations. The results are consistent with the expected behavior of the strong coupling regime and motivate a more comprehensive study to probe the confinement dynamics across the weak and strong coupling regimes.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2025

Introduction: Neuromorphic Materials

The explosive growth in data collection and the need to process it efficiently, as well as the desire to automate increasingly complex tasks in transportation, medical care, manufacturing, security and many other fields have motivated a growing interest in neuromorphic computing. Unlike the binary, transistorbased ON/OFF logic gates and separate logic and memory functionalities employed in digital computing, neuromorphic computing is inspired by animal brains that use interconnected synapses and neurons to perform processing, storage and transmission of information at the same location, while only consuming ~20 W or less of power. Motivated by the brain’s efficiency, adaptability, self-learning and resiliency qualities, neuromorphic computing can be broadly defined as an approach to processing and storing information using hardware and algorithms inspired by models of biological neural systems. Present research in neuromorphic computing encompasses approaches that vary significantly in their degree of neuro-inspiration, from systems that only incorporate features such as asynchronous, event-driven operation or use crossbar arrays of non-volatile memory (NVM) elements to accelerate deep neural networks (DNNs), to designs that embrace the extreme parallelism, sparsity, reconfigurability, adaptability, complexity and stochasticity observed in nervous systems. The term ‘neuromorphic’ computing is often credited to Carver Mead, who in the 1980s investigated Si-based analog electronics to replicate functions of the animal retina. Earlier important advances in this field include the work of Frank Rosenblatt, who proposed the concept of the perceptron, Bernard Widrow, who used this concept to build one of the first analog neural networks, the Adaline and many other researchers (see ref. 6 for an historical perspective on neuromorphic computing). With the recent increase in the use of artificial intelligence and large language models, and rising concerns over the associated energy costs, interest in neuromorphic hardware has expanded rapidly. According to some estimates, driven largely by the drastic growth in the training use of artificial intelligence (AI) models using the current computing architectures, the energy cost of computing is projected to reach the energy supply worldwide by 2045. Furthermore, while this is not a realistic outcome, it means that, if more efficient computing technologies are not developed -- soon -- the world will soon become one where demand for energy and market constraints limit the continued increase of societal access to AI and cloud services from data centers. Data centers used for training and use of these models consume hundreds of terawatt hours of electricity, already past 4% of the US electricity demand.

Circuits↗