Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “matrix models”

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 1,711 records · Page 95

Matrix Multiply Performance of GPUs on Exascale-class HPE/Cray Systems

The computation of dense matrix-matrix products (GEMMs) is central to many modeling and simulation workloads as well as AI/ML deep learning campaigns. In fact, millions of dollars are spent annually on computing GEMMs, and large model training demands are increasing exponentially. Specialized processors such as GPUs are designed to perform well for these operations. However, the performance of GEMMs on GPUs can exhibit complex behaviors depending on many factors, making it challenging to optimize the performance of GEMMs on these processors. In this study we undertake an examination of GEMM performance on several leading GPU models taken from product lines of GPUs to be deployed in forthcoming exascale computing systems. We show results to illustrate the many factors that can affect performance of GEMMs on GPUs. We then present data collected from a large number of test runs for an example GEMM operation to show the dependence behaviors of GEMM rate on matrix dimensions. Finally, we show results from machine learning-based performance models using novel feature engineering methods to fit the measured performance, providing a potential basis for GEMM performance tuning and autotuning methods for GPUs. Recommendations are also given for how to achieve high GEMM performance on modern GPUs.

Melesse Vergara, Veronica↗

Coupled-channels density-matrix approach to low-energy nuclear collision dynamics: A technique for quantifying quantum decoherence effects on reaction observables

The coupled-channels density-matrix technique for nuclear reaction dynamics, which is based on the Liouville-von Neumann equation with Lindblad dissipative terms, is developed with the inclusion of full angular momentum couplings. It allows a quantitative study of the role and importance of quantum decoherence in nuclear scattering. Formulas of asymptotic observables that can reveal effects of quantum decoherence are given. A method for extracting energy-resolved scattering information from the time-dependent density matrix is introduced. As an example, model calculations are carried out for the low-energy collision of the {sup 16}O projectile on the {sup 154}Sm target.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Atomistic model of xenon gas bubble re-solution rate due to thermal spike in uranium oxide

Atomistic simulations are performed to study the response of Xe gas bubbles in UO 2 to ionizing fission products through the thermal spike approximation. A portion of the total electronic stopping power (S e ) is taken as the thermal spike energy through a ratio variable ζ. The thermal spike energy causes extreme melting within the fission track cylindrical region. Molecular dynamics is employed to quantify the probability of a Xe gas atom to be re-solved (re-dissolved) back into the UO 2 matrix. Subsequently, a re-solution model is developed and parametrized as a function of bubble radius (R), off-centered distance (r), and thermal spike energy (ζS e ). The off-centered distance measures the shift of the thermal spike axis from the bubble center. To evaluate the re-solution model, independent fission product yield of U-235 fission due to thermal neutrons (0.0253 eV), taken from the JEFF-3.3 database, is used. The kinetic energy of the fission products is taken from the EXFOR database. Subsequently, the decay of Se over distance for each fission product is simulated. Finally, the evaluated re-solution rate (re-solution probability per second) is presented as a function of bubble radius for a range of ζ.

74 ATOMIC AND MOLECULAR PHYSICS↗

Inelastic response of metal matrix composites under biaxial loading

Theoretical predictions and experimental results were obtained for inelastic response of unidirectional and angle ply composite tubes subjected to axial and torsional loading. The composite material consist of silicon carbide fibers in a titanium alloy matrix. This material is known to be susceptible to fiber matrix interfacial damage. A method to distinguish between matrix yielding and fiber matrix interfacial damage is suggested. Biaxial tests were conducted on the two different layup configurations using an MTS Axial/Torsional load frame with a PC based data acquisition system. The experimentally determined elastic moduli of the SiC/Ti system are compared with those predicted by a micromechanics model. The test results indicate that fiber matrix interfacial damage occurs at relatively low load levels and is a local phenomenon. The micromechanics model used is the method of cells originally proposed by Aboudi. Finite element models using the ABACUS finite element program were used to study end effects and fixture specimen interactions. The results to date have shown good correlation between theory and experiment for response prior to damage initiation.

Lissenden, C. J.↗

Thermal power spectrum in the CFT driven cosmology

We present an overview of the recently suggested cosmological model driven by conformal field theory (CFT) with the initial conditions in the form of the microcanonical density matrix. In particular, we discuss the origin of inflationary stage in this model and a novel feature — the thermal nature of the primordial power spectrum of the CMB anisotropy. The relevant effect of ''temperature of the relic temperature anisotropy'' can be responsible for a thermal contribution to the red tilt of this spectrum, additional to its conventional vacuum component. The amplification of this effect due to recently established a-theorem in CFT is briefly discussed.

79 ASTRONOMY AND ASTROPHYSICS↗

How quantum phases on cylinders approach the two-dimensional limit

We consider the properties of $T$ = $0$ quantum phases of matter - especially superconducting and analogous spin-liquid phases - on infinite cylinders of width $L$ ⊥ and analyze the ways in which the $L$ ⊥ → ∞ (two-dimensional, 2d) limit is approached. This problem is interesting in its own right but is particularly important in the context of extrapolating accessible density matrix renormalization group (DMRG) results on model strongly interacting problems to the desired 2d limit. In conclusion, various methods for drawing firm conclusions about the quantum phases in 2d from relatively small $L$ ⊥ results are illustrated.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

A Micromechanics-Based Damage Model for [+/- Theta/90n]s Composite Laminates

A new damage model based on a micromechanical analysis of cracked [+/- Theta/90n]s laminates subjected to multiaxial loads is proposed. The model predicts the onset and accumulation of transverse matrix cracks in uniformly stressed laminates, the effect of matrix cracks on the stiffness of the laminate, as well as the ultimate failure of the laminate. The model also accounts for the effect of the ply thickness on the ply strength. Predictions relating the elastic properties of several laminates and multiaxial loads are presented.

Mayugo, Joan-Andreu↗

A Micromechanics-Based Damage Model for the Strength Prediction of Composite Laminates

A new damage model based on a micromechanical analysis of cracked [+/-0deg/90deg(sub n)]s laminates subjected to multiaxial loads is proposed. The model predicts the onset and accumulation of transverse matrix cracks in uniformly stressed laminates, the effect of matrix cracks on the stiffness of the laminate, as well as the ultimate failure of the laminate. The model also accounts for the effect of the ply thickness on the ply strength. Predictions relating the elastic properties of several laminates and multiaxial loads are presented.

Camanho, Pedro P.↗

Accelerating Transformer-based Deep Learning Models on FPGAs using Column Balanced Block Pruning

Although Transformer-based language representations achieve state-of-the-art accuracy on various natural language processing (NLP) tasks, the large model size has been challenging the resource constrained computing platforms. Weight pruning, as a popular and effective technique in reducing the number of weight parameters and accelerating the Transformer, has been investigated on GPUs. However, the Transformer acceleration using weight pruning on field-programmable gate array (FPGAs) remains unexplored. This paper investigates the column balanced block-wise pruning on Transformer and designs an FPGA acceleration engine to customize the balanced blockwise matrix multiplication. We implement the Transformer model with proper hardware scheduling, and the experiments show that the Transformer inference on FPGA achieves 10.35 ms latency with the batch size of 32, which is 10.96 × speed up comparing to CPU platform and 2.08 × speed up comparing to GPU platform.

Peng, Hongwu↗

Algorithms for Efficient Computation of Transfer Functions for Large Order Flexible Systems

An efficient and robust computational scheme is given for the calculation of the frequency response function of a large order, flexible system implemented with a linear, time invariant control system. Advantage is taken of the highly structured sparsity of the system matrix of the plant based on a model of the structure using normal mode coordinates. The computational time per frequency point of the new computational scheme is a linear function of system size, a significant improvement over traditional, still-matrix techniques whose computational times per frequency point range from quadratic to cubic functions of system size. This permits the practical frequency domain analysis of systems of much larger order than by traditional, full-matrix techniques. Formulations are given for both open- and closed-loop systems. Numerical examples are presented showing the advantages of the present formulation over traditional approaches, both in speed and in accuracy. Using a model with 703 structural modes, the present method was up to two orders of magnitude faster than a traditional method. The present method generally showed good to excellent accuracy throughout the range of test frequencies, while traditional methods gave adequate accuracy for lower frequencies, but generally deteriorated in performance at higher frequencies with worst case errors being many orders of magnitude times the correct values.

Maghami, Peiman G.↗

Efficient Computation of Closed-loop Frequency Response for Large Order Flexible Systems

An efficient and robust computational scheme is given for the calculation of the frequency response function of a large order, flexible system implemented with a linear, time invariant control system. Advantage is taken of the highly structured sparsity of the system matrix of the plant based on a model of the structure using normal mode coordinates. The computational time per frequency point of the new computational scheme is a linear function of system size, a significant improvement over traditional, full-matrix techniques whose computational times per frequency point range from quadratic to cubic functions of system size. This permits the practical frequency domain analysis of systems of much larger order than by traditional, full-matrix techniques. Formulations are given for both open and closed loop loop systems. Numerical examples are presented showing the advantages of the present formulation over traditional approaches, both in speed and in accuracy. Using a model with 703 structural modes, a speed-up of almost two orders of magnitude was observed while accuracy improved by up to 5 decimal places.

Maghami, Peiman G.↗

Study of the Wigner function at the device boundaries in one-dimensional single- and double-barrier structures

In this work, we compute the Wigner distribution function on one-dimensional devices from wave functions generated by solving the Schroedinger equation. Our goal is to investigate certain issues that we encountered in implementing Wigner transport equation solvers, such as the large discrepancies observed between the boundary conditions and the solution in the neighborhood of the boundaries. By evaluating the Wigner function without solving the Wigner transport equation, we intend to ensure that the actual boundary conditions are consistent with those commonly applied in literature. We study both single- and double-barrier unbiased structures. We use simple potential profiles, so that we can compute the wave functions analytically for better accuracy. We vary a number of structure geometry, material, meshing, and numerical parameters, among which are the contact length, the barrier height, the number of incident wave functions, and the numerical precision used for the computations, and we observe how the Wigner function at the device boundaries is affected. For the double-barrier structures, we look at the density matrix function and we study a model for the device transmission spectrum which helps explain the lobelike artifacts that we observe on the Wigner function.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Transport properties in dilute UN ( X ) solid solutions ( X = Xe , Kr )

Uranium nitride (UN) is a candidate fuel for current GEN III fission reactors, for which it is investigated as an accident-tolerant fuel, as well as for future GEN IV reactors. In this study, we investigate the kinetic properties of gas fission products (Xe and Kr) in UN. Binding and migration energies are obtained using density functional theory, with an added Hubbard correlation to model f electrons, and the occupation matrix control scheme to avoid metastable states. These energies are then used as input for the self-consistent mean field method which enables to determine transport coefficients for vacancy-mediated diffusion of Xe and Kr on the U sublattice. The magnetic ordering of the UN structure is explicitly taken into account, for both energetic and transport properties. Solute diffusivities are compared with experimental measurements and the effect of various parameters on the theoretical model is carefully investigated. We find that kinetic correlations are very strong in this system, and that despite atomic migration anisotropy, macroscopic solute diffusivities show limited anisotropy. Our model indicates that the discrepancy between experimental measurements probably results from different irradiation conditions, and hence different defect concentrations.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Efficacy of the SU(3) scheme for ab initio large-scale calculations beyond the lightest nuclei

We report on the computational characteristics of ab initio nuclear structure calculations in a symmetry-adapted no-core shell model (SA-NCSM) framework. We examine the computational complexity of the current implementation of the SA-NCSM approach, dubbed LSU3shell, by analyzing ab initio results for 6 Li and 12 C in large harmonic oscillator model spaces and SU3-selected subspaces. We demonstrate LSU3shell’s strong-scaling properties achieved with highly-parallel methods for computing the many-body matrix elements. Results compare favorably with complete model space calculations and significant memory savings are achieved in physically important applications. In particular, a well-chosen symmetry-adapted basis affords memory savings in calculations of states with a fixed total angular momentum in large model spaces while exactly preserving translational invariance.

74 ATOMIC AND MOLECULAR PHYSICS↗

Peculiarities of the formation of photonic nanojets by a matrix of dielectric microtoroids

We report the results of theoretical modelling of photonic nanojets (PNJs) formed by laser radiation scattering on a single-layer ordered assembly of dielectric microtoroids placed on the surface of a transparent matrix (silicone film). Using the method of computational electrodynamics (FDTD), the main PNJ parameters (length, width, peak intensity) are analysed under the conditions of mutual influence of the light fields of neighbouring microparticles. It is shown that the main factor affecting the PNJ characteristics under study is the spatial configuration of the radiation-scattering particle, namely, the internal diameter of its cross section. It is found that for certain configurations of toroid placement in a cluster, a PNJ ensemble is implemented with parameters significantly better than those for a single toroid. (paper)

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Efficacy of the SU(3) scheme for ab initio large-scale calculations beyond the lightest nuclei

We report on the computational characteristics of ab initio nuclear structure calculations in a symmetry-adapted no-core shell model (SA-NCSM) framework. We examine the computational complexity of the current implementation of the SA-NCSM approach, dubbed LSU3shell, by analyzing ab initio results for 6 Li and 12 C in large harmonic oscillator model spaces and SU(3)-selected subspaces. We demonstrate LSU3shell's strong-scaling properties achieved with highly-parallel methods for computing the many-body matrix elements. Results compare favorably with complete model space calculations and signi cant memory savings are achieved in physically important applications. In particular, a well-chosen symmetry-adapted basis a ords memory savings in calculations of states with a fixed total angular momentum in large model spaces while exactly preserving translational invariance.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗