SEARCH · Engineering Papers
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.
The purpose of this microbenchmark is to provide a means to explore different programming methodologies using a simple, but not trivial, mathematical kernel. The kernel is based on the mult_su3_nn() SU(3) matrix-matrix multiply routine in the MILC Lattice Quantum Chromodynamics(LQCD) code.
The purpose of this microbenchmark is to provide a means to explore different programming methodologies using a simple, but not trivial, mathematical kernel. The kernel is based on the matrix vector product in the MFDn nuclear configuration interaction code's LOBPCG eigensolver. This particular mathematical operation is a key element in the solution of systems of equations and block eigensolvers.
We would like to create an open source repository in LANL's github on code written in Julia, in which we implement and extend the data-driven Mori-Zwanzig method for extracting large-scale spatio-temporal structures from data, which we call MZMD. This method is an extension of Dynamic Mode Decomposition (DMD) in which Mori-Zwanzig memory kernels are included into the associated companion matrix. In the code we would like to release, we apply MZMD to a flow over a cylinder with Reynolds number 100 rather than the much larger data set used in the associated manuscript. DMD is used extensively in the fluid dynamics community mainly for extracting large scale spatio-temporal structures (patters) from flow data. This is useful for understanding the key mechanisms that generate certain complex dynamical process relevant in engineering design. In MZMD, we improve upon DMD by adding the Mori-Zwanzig memory kernels, and show this improvement is especially important in strongly nonlinear regions of the flow.
The sea-ice-albedo feedback (SIAF) is the product of the ice sensitivity (IS) – how much the surface albedo in sea-ice regions changes as the planet warms– and the radiative sensitivity (RS) – how much the top of atmosphere radiation changes as the surface albedo changes. We demonstrate that the RS calculated from radiative kernels in climate models is reproduced from calculations using the “approximate partial radiative perturbation” method that uses the climatological radiative fluxes at the top of atmosphere and the assumption that the atmosphere is isotropic to shortwave radiation. This method facilitates the comparison of RS from satellite-based estimates of climatological radiative fluxes with RS estimates across a full suite of coupled climate models and, thus, allows model evaluation of a quantity important in characterizing the climate impact of sea ice concentration changes. The satellite based RS is within the model range of RS that differs by a factor of two across climate models in both the Arctic and Southern Ocean. Observed trends in Arctic sea ice are used to estimate IS which, in conjunction with the satellite-based RS yields an SIAF of 0.16 ± 0.04 W m-2 K-1. This Arctic SIAF estimate suggests a modest amplification of future global surface temperature change by approximately 14% relative to a climate system with no SIAF. We calculate the global albedo feedback in climate models using model specific RS and IS and find a model mean feedback parameter of 0.37 W m-2 K-1 which is 40% larger than the IPCC AR5 estimate based on using RS calculated from radiative kernel calculations in a single climate model.
A single-channel (56.7 GHz) scanning radiometer was deployed in August 2022 at the Atmospheric Radiation Measurement (ARM) North Slope of Alaska site near Utqiaġvik. The radiometer is designed to provide temperature profiles between 0 and 1 km every 5 min. Averaging kernels show that this single-channel radiometer, taking observations at 10 discrete elevation angles, yields approximately the same information as a seven-channel V-band radiometer scanning three elevation angles. The instrument is able to reproduce the occurrence of temperature inversions between the surface and 1 km and their strength showing a correlation of 0.85, bias of −0.6 K, and slope of 1.04 with respect to radiosondes. Uncertainty in the inversion base height varies from 50 m near the surface to ∼300 m above 0.4 km when compared with radiosondes. Conversely, the inversion top height is overestimated and has higher uncertainty due to the degrading effects of the averaging kernels on the vertical resolution of the retrievals. Thanks to the high temporal resolution of the retrievals, the diurnal cycle of boundary layer temperature was evaluated showing that the radiometer can capture some aspects of the boundary layer thermal structure. The present analysis provides an overview of the capabilities of this simple observing configuration for selected applications.
Performance tests and analyses are critical to effective high-performance computing software development and are central components in the design and implementation of computational algorithms for achieving faster simulations on existing and future computing architectures for large-scale application problems. In this article, we explore performance and space-time trade-offs for important compute-intensive kernels of large-scale numerical solvers for partial differential equations (PDEs) that govern a wide range of physical applications. We consider a sequence of PDE-motivated bake-off problems designed to establish best practices for efficient high-order simulations across a variety of codes and platforms. We measure peak performance (degrees of freedom per second) on a fixed number of nodes and identify effective code optimization strategies for each architecture. In addition to peak performance, we identify the minimum time to solution at 80% parallel efficiency. The performance analysis is based on spectral and p-type finite elements but is equally applicable to a broad spectrum of numerical PDE discretizations, including finite difference, finite volume, and h-type finite elements.
On the path to exascale the landscape of computer device architectures and corresponding programming models has become much more diverse. While various low-level performance portable programming models are available, support at the application level lacks behind. To address this issue, we present the performance portable block-structured adaptive mesh refinement (AMR) framework Parthenon, derived from the well-tested and widely used Athena++ astrophysical magnetohydrodynamics code, but generalized to serve as the foundation for a variety of downstream multi-physics codes. Parthenon adopts the Kokkos programming model, and provides various levels of abstractions from multidimensional variables, to packages defining and separating components, to launching of parallel compute kernels. Parthenon allocates all data in device memory to reduce data movement, supports the logical packing of variables and mesh blocks to reduce kernel launch overhead, and employs one-sided, asynchronous MPI calls to reduce communication overhead in multi-node simulations. Using a hydrodynamics miniapp, we demonstrate weak and strong scaling on various architectures including AMD and NVIDIA GPUs, Intel and AMD x86 CPUs, IBM Power9 CPUs, as well as Fujitsu A64FX CPUs. At the largest scale on Frontier (the first TOP500 exascale machine), the miniapp reaches a total of 1.7 × 10 13 zone-cycles/s on 9216 nodes (73,728 logical GPUs) at [Formula: see text] weak scaling parallel efficiency (starting from a single node). In combination with being an open, collaborative project, this makes Parthenon an ideal framework to target exascale simulations in which the downstream developers can focus on their specific application rather than on the complexity of handling massively-parallel, device-accelerated AMR.
We present hipBone, an open-source performance-portable proxy application for the Nek5000 (and NekRS) computational fluid dynamics applications. HipBone is a fully GPU-accelerated C++ implementation of the original NekBone CPU proxy application with several novel algorithmic and implementation improvements which optimize its performance on modern fine-grain parallel GPU accelerators. Our optimizations include a conversion to store the degrees of freedom of the problem in assembled form in order to reduce the amount of data moved during the main iteration and a portable implementation of the main Poisson operator kernel. We demonstrate near-roofline performance of the operator kernel on three different modern GPU accelerators from two different vendors. We present a novel algorithm for splitting the application of the Poisson operator on GPUs which aggressively hides MPI communication required for both halo exchange and assembly. Our implementation of nearest-neighbor MPI communication then leverages several different routing algorithms and GPU-Direct RDMA capabilities, when available, which improves scalability of the benchmark. We demonstrate the performance of hipBone on three different clusters housed at Oak Ridge National Laboratory, namely, the Summit supercomputer and the Frontier early-access clusters, Spock and Crusher. Our tests demonstrate both portability across different clusters and very good scaling efficiency, especially on large problems.
We investigate the solution of low-rank matrix approximation problems using the truncated singular value decomposition (SVD). For this purpose, we develop and optimize graphics processing unit (GPU) implementations for the randomized SVD and a blocked variant of the Lanczos approach. Our work takes advantage of the fact that the two methods are composed of very similar linear algebra building blocks, which can be assembled using numerical kernels from existing high-performance linear algebra libraries. Furthermore, the experiments with several sparse matrices arising in representative real-world applications and synthetic dense test matrices reveal a performance advantage of the block Lanczos algorithm when targeting the same approximation accuracy.
Over the course of interactions with various application teams, the need for batched sparse linear algebra functions has emerged in order to make more efficient use of the GPUs for many small and sparse linear algebra problems. In this paper, we present our recent work on a batched sparse direct solver for GPUs. The sparse LU factorization is computed by the levels of the elimination tree, leveraging the batched dense operations at each level and a new batched Scatter GPU kernel. The sparse triangular solve is computed by the level sets of the directed acyclic graph (DAG) of the triangular matrix. Batched operations overcome the large overhead associated with launching many small kernels. For medium sized matrix batches with not-so-small bandwidth, using an NVIDIA A100 GPU, our new batched sparse direct solver is orders of magnitude faster than a batched banded solver and uses less than one-tenth of the memory.
The Portable Extensible Toolkit for Scientific Computation (PETSc) library provides scalable solvers for nonlinear time-dependent differential and algebraic equations and for numerical optimization via the Toolkit for Advanced Optimization (TAO). PETSc is used in dozens of scientific fields and is an important building block for many simulation codes. During the U.S. Department of Energy’s Exascale Computing Project, the PETSc team has made substantial efforts to enable efficient utilization of the massive fine-grain parallelism present within exascale compute nodes and to enable performance portability across exascale architectures. We recap some of the challenges that designers of numerical libraries face in such an endeavor, and then discuss the many developments we have made, which include the addition of new GPU backends, features supporting efficient on-device matrix assembly, better support for asynchronicity and GPU kernel concurrency, and new communication infrastructure. In conclusion, we evaluate the performance of these developments on some pre-exascale systems as well as the early exascale systems Frontier and Aurora, using compute kernel, communication layer, solver, and mini-application benchmark studies, and then close with a few observations drawn from our experiences on the tension between portable performance and other goals of numerical libraries.
This paper investigates the effect of laminar-to-turbulent flame transition modeling on the prediction of cycle-to-cycle variations (CCVs) in large eddy simulation (LES) of spark-ignition (SI) engines. A laminar-to-turbulent flame transition model that describes the non-equilibrium sub-filter flame speed evolution during an early stage of flame kernel growth is developed. In the present model, the flame transition is characterized by the flame kernel size at which the flame transition ends, defined here as the flame transition scale. The proposed model captures the effects that variations in a turbulent flow field have on the evolution of early-stage burning rates, through variations in the flame transition scale. The proposed flame transition model is combined with the front propagation formulation (FPF) method and a spark-ignition model to predict CCVs in a gasoline direct injection SI engine. It is found that multi-cycle LES with the proposed flame transition model reproduces experimentally-observed CCVs satisfactorily. When the transition model is not considered or when variations in the transition process are neglected, CCVs are significantly under-predicted for the case considered here. These results indicate the importance of modeling the laminar-to-turbulent flame transition and the effect of turbulence on the transition process, when predicting CCVs, under certain engine conditions. The LES results are also used to analyze sources for variations in the flame transition. It is found, for the present engine case, that the most important source is the cycle-to-cycle variation in the turbulence dissipation rate, which is used to measure the strength of turbulence in the proposed model, near a spark plug. The large-scale velocity field and the variations of the laminar flame speed due to the mixture composition and thermal stratification are also found to be important factors to contribute to the variations in the flame transition.
Formulations of Neumann-type boundary conditions for boundary value problems in the nonlocal framework are beset with difficulties, some related to the choice of a proper scaling. Here we identify a space-dependent scaling for a nonlocal Neumann operator, for which we prove linear in δ (δ being the radius for the support for the kernel) convergence of the Neumann operator and $\mathcal{O}$(δ 2 ) convergence of solutions to their classical counterparts. The pointwise-like convergence of the nonlocal normal operator is cast as a new type of two-scale operator-point convergence, which we call condensated convergence . The results hold for general integrable kernels, a setting which is favored in numerical simulations. We support this analysis with numerical convergence studies using a piecewise linear discontinuous Galerkin discretization and show an $\mathcal{O}$(δ 2 ) rate of convergence of solutions, also exhibiting an $\mathcal{O}$(h 2 ) convergence, where h is the mesh size.
This work investigates and applies machine learning paradigms seldom seen in analytical spectroscopy for quantification of gallium in cerium matrices via processing of laser-plasma spectra. Ensemble regressions, support vector machine regressions, Gaussian kernel regressions, and artificial neural network techniques are trained and tested on cerium-gallium pellet spectra. A thorough hyperparameter optimization experiment is conducted initially to determine the best design features for each model. The optimized models are evaluated for sensitivity and precision using the limit of detection (LoD) and root mean-squared error of prediction (RMSEP) metrics, respectively. Gaussian kernel regression yields the superlative predictive model with an RMSEP of 0.33% and an LoD of 0.015% for quantification of Ga in a Ce matrix. This study concludes that these machine learning methods could yield robust prediction models for rapid quality control analysis of plutonium alloys.
3D object recognition is one of the most popular areas of study in computer vision. Many of the more recent algorithms focus on indoor point clouds, classifying 3D geometric objects, and segmenting outdoor 3D scenes. One of the challenges of the classification pipeline is finding adequate and accurate training data. Hence, this article seeks to evaluate the accuracy of a synthetically generated data set called SynthCity, tested on two mobile laser-scan data sets. Varying levels of noise were applied to the training data to reflect varying levels of noise in different scanners. The chosen deep-learning algorithm was Kernel Point Convolution, a convolutional neural network that uses kernel points in Euclidean space for convolution weights.
The heat kernel and quasinormal mode methods of computing 1-loop partition functions of spin s s fields on hyperbolic quotient spacetimes \mathbb{H}^{3}/\mathbb{Z} ℍ 3 / ℤ are related via the Selberg zeta function. We extend that analysis to thermal \text{AdS}_{2n+1} AdS 2 n + 1 backgrounds, with quotient structure \mathbb{H}^{2n+1}/\mathbb{Z} ℍ 2 n + 1 / ℤ . Specifically, we demonstrate the zeros of the Selberg function encode the normal mode frequencies of spin fields upon removal of non-square-integrable modes. With this information we construct the 1-loop partition functions for symmetric transverse traceless tensors in terms of the Selberg zeta function and find exact agreement with the heat kernel method.
We propose an intuitive, machine-learning approach to multiparameter inference, dubbed the InferoStatic Networks (ISN) method, to model the score and likelihood ratio estimators in cases when the probability density can be sampled but not computed directly. The ISN uses a backend neural network that models a scalar function called the inferostatic potential \varphi φ . In addition, we introduce new strategies, respectively called Kernel Score Estimation (KSE) and Kernel Likelihood Ratio Estimation (KLRE), to learn the score and the likelihood ratio functions from simulated data. We illustrate the new techniques with some toy examples and compare to existing approaches in the literature. We mention en passant some new loss functions that optimally incorporate latent information from simulations into the training procedure.
The small- x x quark helicity evolution equations at double-logarithmic order, with the kernel \sim\alpha_s\ln^2(1/x) ∼ α s ln 2 ( 1 / x ) , have been derived previously. In this work, we derive the single-logarithmic corrections to the equations, to order \alpha_s\ln(1/x) α s ln ( 1 / x ) of the evolution kernel. The new equations include the effects of the running coupling and the unpolarized small- x x evolution, both of which are parametrically significant at single-logarithmic order. The large- N_c N c and large- N_c\ N_f N c N f approximations to the equation are computed. (Here, N_c N c and N_f N f are the numbers of quark colors and flavors, respectively.) Their solutions will provide more precise estimates of the quark helicity distribution at small x x , contributing to the resolution of the proton spin puzzle.