Search NASA⌕ Search

SEARCH · Search NASA

Results for “Speedup model”

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 OSTI2025

A GPU Accelerated Mixed‐Precision Finite Difference Informed Random Walker (FDiRW) Solver for Strongly Inhomogeneous Diffusion Problems

In nature, many complex multi‐physics coupling problems exhibit significant diffusivity inhomogeneity, where one process occurs several orders of magnitude faster than others temporally. Simulating rapid diffusion alongside slower processes demands intensive computational resources due to the necessity for small time steps. To address these computational challenges, we have developed an efficient numerical solver named Finite Difference informed Random Walker (FDiRW). In this study, we propose a GPU‐accelerated, mixed‐precision configuration for the FDiRW solver to maximize efficiency through GPU multi‐threaded parallel computation and lower precision computation. Numerical evaluation results reveal that the proposed GPU‐accelerated mixed‐precision FDiRW solver can achieve a 117× speedup over the CPU baseline, while an additional 1.75× speedup is achieved by employing lower precision GPU computation. Notably, for large model sizes, the GPU‐accelerated mixed‐precision FDiRW solver demonstrates strong scaling with the number of nodes used in simulation. When simulating radionuclide absorption processes by porous wasteform particles with a medium‐sized model of 192 × 192 × 192, this approach reduces the total computational time to 10 min, enabling the simulation of larger systems with strongly inhomogeneous diffusivity.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Shortened Duration of Global Warming Slowdowns with Elevated Greenhouse Gas Emissions

Continuous emissions of anthropogenic greenhouse gases (GHGs) and aerosols in the last 160 years have resulted in an increasing trend of global mean surface temperatures (GMSTs). Due to interactions with natural variability, rates of the combined anthropogenically and naturally induced warming trends are characterized by significant slowdowns and speedups on decadal timescales. In this work, by analyzing observed and model-simulated data, we investigate how the duration of these episodes will change with different strengths of GHG and aerosol forcing. We found that the duration of warming slowdowns can be more than 30 yr with a slower rate of anthropogenic emissions but would shorten to about 5 yr with a higher one. This duration reduction depends on both the magnitude of the climate response to anthropogenic forcing and the strength of the internal variability. Moreover, the warming slowdowns can still occur even towards the end of this century under high emissions scenarios but with significantly shortened duration.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

A machine learning approach for increased throughput of density functional theory substitutional alloy studies

In this study, a machine learning-based technique is developed to reduce the computational cost required to explore large design spaces of substitutional alloys. The first advancement is based on a neural network (NN) approach to predict the initial position of minority and majority ions prior to DFT relaxation. The second advancement is to allow the NN to predict the total energy for every possibility minority ion position and select the most stable configuration in the absence of relaxing each trial minority configuration. A bismuth oxide materials system, (Bi x La y Yb z ) 2 MoO 6 , was used as the model system to demonstrate the developed methods and quantify the resulting computational speedup. Compared to a brute force method that requires the calculation of every permutation of minority configuration and subsequent DFT relaxation, a 1.3× speedup was realized if the NN predicted the initial configuration of ions prior to relaxation. Implementation of the second advancement allowed the NN to predict the total energy for all possible trial configurations and downselect the most stable configurations prior to relaxation, resulting in a speedup of approximately 37×. Validation was done by comparing position and energy between the NN and DFT predictions. A maximum position vector mean squared error (MSE) of 1.6 × 10 -2 and a maximum energy MSE of 2.3 × 10 -7 was predicted for the worst case configuration. This method demonstrates a significant computational speedup, which has the potential for even greater computational savings for larger compositional design spaces.

36 MATERIALS SCIENCE↗
DOE OSTI2014

High speed finite element simulations on the graphics card

A software package is developed to perform explicit time domain finite element simulations of ultrasonic propagation on the graphical processing unit, using Nvidia’s CUDA. Of critical importance for this problem is the arrangement of nodes in memory, allowing data to be loaded efficiently and minimising communication between the independently executed blocks of threads. The initial stage of memory arrangement is partitioning the mesh; both a well established ‘greedy’ partitioner and a new, more efficient ‘aligned’ partitioner are investigated. A method is then developed to efficiently arrange the memory within each partition. The technique is compared to a commercial CPU equivalent, demonstrating an overall speedup of at least 100 for a non-destructive testing weld model.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2021

Quantum computing without quantum computers: Database search and data processing using classical wave superposition

Quantum computers are proven to be more efficient at solving a specific class of problems compared to traditional digital computers. Superposition of states and quantum entanglement are the two key ingredients that make quantum computing so powerful. However, not all quantum algorithms require quantum entanglement (e.g., search through an unsorted database). Is it possible to utilize classical wave superposition to speed up database searching as much as by using quantum computers? There were several attempts to mimic quantum computers using classical waves. It was concluded that the use of classical wave superposition comes with the cost of an exponential increase in resources. In this work, we consider the feasibility of building classical wave-based devices able to provide fundamental speedup over digital counterparts without the exponential overhead. We present experimental data on database searching through a magnetic database using spin wave superposition. The results demonstrate the same speedup as expected for quantum computers. Also, we present examples of numerical modeling demonstrating classical wave interference for period finding. This approach may not compete with quantum computers with efficiency but outperform classical digital computers. We argue that classical wave-based devices can perform some of the quantum algorithms with the same efficiency as quantum computers as long as quantum entanglement is not required.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
NASA NTRS2016

A Computationally-Efficient Inverse Approach to Probabilistic Strain-Based Damage Diagnosis

This work presents a computationally-efficient inverse approach to probabilistic damage diagnosis. Given strain data at a limited number of measurement locations, Bayesian inference and Markov Chain Monte Carlo (MCMC) sampling are used to estimate probability distributions of the unknown location, size, and orientation of damage. Substantial computational speedup is obtained by replacing a three-dimensional finite element (FE) model with an efficient surrogate model. The approach is experimentally validated on cracked test specimens where full field strains are determined using digital image correlation (DIC). Access to full field DIC data allows for testing of different hypothetical sensor arrangements, facilitating the study of strain-based diagnosis effectiveness as the distance between damage and measurement locations increases. The ability of the framework to effectively perform both probabilistic damage localization and characterization in cracked plates is demonstrated and the impact of measurement location on uncertainty in the predictions is shown. Furthermore, the analysis time to produce these predictions is orders of magnitude less than a baseline Bayesian approach with the FE method by utilizing surrogate modeling and effective numerical sampling approaches.

Warner, James E.↗
DOE OSTI2025

Predicting Flow in Fracture Networks With Quantum Algorithms

Uncertainty quantification plays a crucial role in the modeling of subsurface flow. For instance, uncertainties in the properties of geologic fracture networks significantly impact flow, requiring numerous simulations to accurately estimate quantities of interest. However, each simulation is computationally expensive because it requires solving a large linear system to capture features that involve both small and large fractures. An example is in percolation, where the interaction of many small fractures (which cumulatively can have a large surface area) with the rock matrix must be modeled precisely. Quantum computing is an emerging tool with the potential to address this issue. Quantum algorithms offer a significant speedup in solving linear systems, achieving efficiencies that are challenging to match with classical approaches. These classical approaches include direct solvers, such as LU decomposition, and iterative methods, notably preconditioned conjugate gradient, commonly used in subsurface modeling to solve large sparse systems. However, applying quantum algorithms to geologic fracture flow requires careful attention to algorithmic and problem-specific constraints to fully realize this quantum advantage. In this work we describe a quantum algorithm for generalized Monte Carlo applications with a quadratic speedup over the classical approaches which can be combined with the quantum speedup, currently under investigation, for solving quantum linear systems for subsurface flow. We show that for quantum algorithms the computational cost of estimating a quantity of interest for a statistical ensemble of networks is roughly the same as that of a single realization, essentially implying that one can get uncertainty quantification for free.

58 GEOSCIENCES↗
DOE OSTI2021

Theory of Trotter Error with Commutator Scaling

The Lie-Trotter formula, together with its higher-order generalizations, provides a simple approach to decomposing the exponential of a sum of operators. Despite significant effort, the error scaling of such product formulas remains poorly understood. We develop a theory of Trotter error that overcomes the limitations of truncating the Baker-Campbell-Hausdorff expansion. Our analysis directly exploits the commutativity of operator summands, producing tighter error bounds for both real- and imaginary-time evolutions. Whereas previous work achieves similar goals for systems with geometric locality or Lie-algebraic structure, our approach holds in general. We give a host of improved algorithms for digital quantum simulation and quantum Monte Carlo methods, nearly matching or even outperforming the best previous results. Our applications include: (i) a simulation of second-quantized plane-wave electronic structure, nearly matching the interaction-picture algorithm of Low and Wiebe; (ii) a simulation of $k$-local Hamiltonians almost with induced one-norm scaling, faster than the qubitization algorithm of Low and Chuang; (iii) a simulation of rapidly decaying power-law interactions, outperforming the Lieb-Robinson-based approach of Tran et al.; (iv) a hybrid simulation of clustered Hamiltonians, dramatically improving the result of Peng, Harrow, Ozols, and Wu; and (v) quantum Monte Carlo simulations of the transverse field Ising model and quantum ferromagnets, tightening previous analyses of Bravyi and Gosset. We obtain further speedups using the fact that product formulas can preserve the locality of the simulated system. Specifically, we show that local observables can be simulated with complexity independent of the system size for power-law interacting systems, which implies a Lieb-Robinson bound nearly matching a recent result of Tran et al. Our analysis reproduces known tight bounds for first- and second-order formulas. We further investigate the tightness of our bounds for higher-order formulas. For quantum simulation of a one-dimensional Heisenberg model with an even-odd ordering of terms, our result overestimates the complexity by only a factor of $5$. Our bound is also close to tight for power-law interactions and other orderings of terms. This suggests that our theory can accurately characterize Trotter error in terms of both the asymptotic scaling and the constant prefactor.

quantum computing, numerical analysis↗
NASA NTRSDate not supplied

Quantum-Accelerated Distributed Algorithms for Approximate Steiner Trees and Directed Minimum Spanning Trees

We present two algorithms in the Quantum CONGEST-CLIQUE model of distributed computation that succeed with high probability; One for producing an approximately optimal Steiner Tree, and one for producing an exact Minimum Directed Spanning tree. These use O(n1/4) rounds of communication and O(n9/4) messages, leading to a quantum speedup in round and message complexity compared to any known algorithms in the classical CONGEST-CLIQUE model (vs O(n1/3) and O(n7/3)). At a high level, we achieve these results by combining classical algorithms with fast quantum subroutines. Further, these problems can not be sped up in the CONGEST (non-clique) setting, and we characterize the constants involved.

Phillip Kerger↗
DOE OSTI2023

Uncertainty-aware molecular dynamics from Bayesian active learning for phase transformations and thermal transport in SiC

Abstract Machine learning interatomic force fields are promising for combining high computational efficiency and accuracy in modeling quantum interactions and simulating atomistic dynamics. Active learning methods have been recently developed to train force fields efficiently and automatically. Among them, Bayesian active learning utilizes principled uncertainty quantification to make data acquisition decisions. In this work, we present a general Bayesian active learning workflow, where the force field is constructed from a sparse Gaussian process regression model based on atomic cluster expansion descriptors. To circumvent the high computational cost of the sparse Gaussian process uncertainty calculation, we formulate a high-performance approximate mapping of the uncertainty and demonstrate a speedup of several orders of magnitude. We demonstrate the autonomous active learning workflow by training a Bayesian force field model for silicon carbide (SiC) polymorphs in only a few days of computer time and show that pressure-induced phase transformations are accurately captured. The resulting model exhibits close agreement with both ab initio calculations and experimental measurements, and outperforms existing empirical models on vibrational and thermal properties. The active learning workflow readily generalizes to a wide range of material systems and accelerates their computational understanding.

36 MATERIALS SCIENCE↗
DOE OSTI2014

Approximate Hessian for accelerating ab initio structure relaxation by force fitting

We present a method to approximate the Hessian matrix of the Born-Oppenheimer energy landscape by using a simple force field model whose parameters are fitted to on-the-flight ab initio results. The inversed Hessian matrix is used as the preconditioner of conjugate gradient algorithms to speed up the atomic structure relaxation, resulting in a speedup factor of 2 to 5 on systems of bulk, slab, sheets, and atomic clusters. Because the force field model employed is simple and general, the parameter fitting is straightforward; the method is applicable to a variety of complicated systems for minimum structure relaxation. In the metal cluster new structure search, the new method yields better structures than the one obtained before with conventional algorithms.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Analysis of GPU Data Access Patterns on Complex Geometries for the D3Q19 Lattice Boltzmann Algorithm

GPU performance of the lattice Boltzmann method (LBM) depends heavily on memory access patterns. When implemented with GPUs on complex domains, typically, geometric data is accessed indirectly and lattice data is accessed lexicographically. Although there are a variety of other options, no study has examined the relative efficacy between them. Here, we examine a suite of memory access schemes via empirical testing and performance modeling. We find strong evidence that semi-direct is often better suited than the more common indirect addressing, providing increased computational speed and reducing memory consumption. For the layout, we find that the Collected Structure of Arrays (CSoA) and bundling layouts outperform the common Structure of Array layout; on V100 and P100 devices, CSoA consistently outperforms bundling, however the relationship is more complicated on K40 devices. When compared to state-of-the-art practices, our recommendations lead to speedups of 10–40 percent and reduce memory consumption up to 17 percent. Using performance modeling and computational experimentation, we determine the mechanisms behind the accelerations. We demonstrate that our results hold across multiple GPUs on two leadership class systems, and present the first near-optimal strong results for LBM with arterial geometries run on GPUs.

42 ENGINEERING↗
DOE OSTI2023

Accelerating the Lagrangian Particle Tracking in Hydrologic Modeling to Continental-Scale

Unprecedented climate change and anthropogenic activities have induced increasing ecohydrological problems, which have motivated the development of large-scale modeling for solutions. Water age/quality is as important as water quantity for understanding the water cycle. However, current scientific progress in tracking water parcels at large-scale with high spatiotemporal resolutions is far behind that in water balance/quantity owing to the lack of powerful tools. EcoSLIM is a particle tracking model that works with the hydrologic model ParFlow-CLM, which couples surface-subsurface hydrology with land surface processes. Here, we demonstrate a parallel framework to accelerate EcoSLIM to continental-scale on a distributed, multi-GPU platform with CUDA-Aware MPI. In tests from catchment-, to regional-, and then to continental-scale using 25-million to 1.6-billion particles, EcoSLIM shows significant speedup and excellent parallel performance. The parallel framework is portable to atmospheric and oceanic particle tracking models, where parallelization is inadequate and a standard parallel framework is absent. Parallelized EcoSLIM is a promising tool to accelerate our understanding of the terrestrial water cycle and the upscaling of subsurface hydrology to Earth system models.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS1990

Optimal pipelining

An effort is made to characterize the tradeoffs and overheads limiting the speedup potential theoretically projected for pipeline-incorporating computer architectures, using a mathematical model of the roles played by the various parameters. Pipeline optimization proceeds by a partitioning of the pipeline into an optimum number of segments so that maximization of throughput is obtained. Inferences are drawn from the model, and potential improvements to it are identified. Substantial agreement is obtained with Kunkel and Smith's (1986) CRAY-1S simulations of pipelining.

Dubey, Pradeep K.↗
DOE OSTI2018

A Novel In Situ Flamelet Tabulation Methodology for the Representative Interactive Flamelet Model

A priori flamelet tabulation methods have been used extensively to model premixed and non-premixed combustion in Reynolds Averaged Navier-Stokes (RANS) and Large Eddy Simulations (LES). Pre-tabulation of flamelet libraries have led to significant cost reduction of CFD simulations. However, these tabulation methods rely on certain simplifying assumptions. The dimensionality of the flamelet manifolds and size of flamelet libraries are also constrained due to memory and computational limitations. This limits the applicability and predictive capability of flamelet models. The traditional representative interactive flamelet (RIF) approach with multiple flamelets on the other hand involves online solution of flamelet equations and presumed PDF integration at each computational cell making it attractive but computationally expensive and not suitable for LES with large cell counts. The current work discusses the development of a novel methodology that combines the advantages of an online flamelet solver along with the computational efficacy of tabulated models. The novel in situ tabulation approach consists of creating a flamelet table at each computational time step using the unsteady flamelet equations. The flamelets are then integrated in real time based on their respective scalar dissipation rate histories. This approach is capable of including history effects and unsteady chemical kinetic effects as it involves the online solution of unsteady flamelets. Implementation and comprehensive validation of the new framework is carried out against gas jet flames as well as spray flames at high pressures. The approach is first validated for a partially premixed methane gas-jet flame in a RANS framework. The temperature and species mass fractions are compared against experimental results at different locations of a lifted flame. The approach is able to capture the transient flame characteristics. The model is then validated against an n -dodecane spray flame at high pressures in a LES framework with a 103 species n -dodecane chemistry mechanism. The model is able to capture ignition delays and flame liftoff of the n -dodecane spray over varying ambient temperature conditions. Due to the offsetting of presumed PDF integration cost, the proposed framework is shown to be faster than the traditional RIF approach and the speedup over RIF increases with an increase in the cell count. Furthermore, this framework enables use of the RIF model for large cell count LES simulations.

42 ENGINEERING↗
NASA NTRS2018

Mechanism of Quantum Speedup in Novel Population Transfer Protocol for Binary Optimization Problems

We consider a novel quantum population transfer protocol to solve binary optimization problems that exploits quantum many-body dynamics in the delocalized regime. Hard optimization problems are characterized by energy landscape with a large number of local minima separated by large Hamming distances which scale with the problem size. This landscape gives rise to an interesting computational primitive: given an initial bit-string, we are to produce other bit-strings within certain narrow range of energies around the initial state. We consider a specific model we call "impurity band": a system of n qubits in a transverse field, where a number of bitstrings $M<<2^n$ selected at random are assigned random energies distributed in a narrow window of width $W<<1$ around the mean energy $-n$. We demonstrate the existence of the many-body delocalized regime in this model when the spectrum of the model splits into many-body minibands, and a typical eigenstate wave function is a superposition of peaks centered at a large number of local minima. The typical width of the minibands in energy determines the efficiency of the population transfer protocol. We demonstrate theoretically that the population transfer protocol achieves Grover type speedup in the unstructured impurity band model.

Kechedzhi, Kostyantyn↗
DOE OSTI2025

RingX: Scalable Parallel Attention for Long-Context Learning on HPC

The attention mechanism has become foundational for remarkable AI breakthroughs since the introduction of the Transformer, driving the demand for increasingly longer context to power frontier models such as large-scale reasoning language models and high-resolution image/video generators. However, its quadratic computational and memory complexities present substantial challenges. Current state-of-the-art parallel attention methods, such as ring attention, are widely adopted for long-context training but utilize a point-to-point communication strategy that fails to fully exploit the capabilities of modern HPC network architectures. In this work, we propose ringX, a scalable family of parallel attention methods optimized explicitly for HPC systems. By enhancing workload partitioning, refining communication patterns, and improving load balancing, ringX achieves up to 3.4 × speedup compared to conventional ring attention on the Frontier supercomputer. Optimized for both bi-directional and causal attention mechanisms, ringX demonstrates its effectiveness through training benchmarks of a Vision Transformer (ViT) on a climate dataset and a Generative Pre-Trained Transformer (GPT) model, Llama3 8B. Our method attains an end-to-end training speedup of approximately 1.5 × in both scenarios. To our knowledge, the achieved 38% model FLOPs utilization (MFU) for training Llama3 8B with a 1M-token sequence length on 4,096 GPUs represents one of the highest training efficiencies reported for long-context learning on HPC systems. Our code implementation is available at https://github.com/jqyin/ringX-attention.

Yin, Junqi [ORNL] (ORCID:0000000338435520)↗
DOE OSTI2025

A physics-constrained neural ordinary differential equations approach for robust learning of stiff chemical kinetics

The high computational cost associated with solving for detailed chemistry poses a significant challenge for predictive computational fluid dynamics (CFD) simulations of turbulent reacting flows. While deep learning techniques have been explored to develop faster surrogate models, they often fail to integrate reliably with CFD solvers. This instability arises because traditional deep learning approaches optimize for training error without ensuring compatibility with ordinary differential equation (ODE) solvers, resulting in accumulation of errors over time. Recently, neuralODE (NODE) based approaches have been shown to be a promising technique to emulate and accelerate detailed chemistry computations. Here, in the present work, we extend this NODE framework for stiff chemical kinetics by incorporating mass conservation constraints directly into the loss function during training. This ensures that the total mass as well as the individual elemental species masses are conserved in an a-posteriori manner. Proof-of-concept studies are performed with the novel physics-constrained NODE (PC-NODE) approach for homogeneous autoignition of hydrogen-air mixture over a range of composition and thermodynamic conditions. It is demonstrated that the PC-NODE framework not only improves the physical consistency of the resulting data-driven model with respect to mass conservation criteria, but also improves training efficiency. PC-NODE is shown to achieve 2–100× speedup relative to the hydrogen-air detailed chemical mechanism depending on the type of the ODE solver (implicit or explicit) used during autoregressive inference tests. Lastly, a-posteriori studies are performed wherein the trained PC-NODE model is coupled with a CFD solver. It is shown that higher accuracy is achieved with PC-NODE relative to the purely data-driven NODE approach. Moreover, PC-NODE also exhibits robustness and generalizability to unseen initial conditions from within (interpolative capability) as well as outside (extrapolative capability) the training regime.

computational combustion↗