Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Error Modeling”

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 1,441 records · Page 80

Multi-Scale Modeling of Thermal-Fluid Phenomena Related to Loss of Forced Circulation Transient in HTGRs

Under the support of DOE-NE’s Nuclear Energy Advanced Modeling and Simulation (NEAMS), an effort had been pursued to support High Temperature Gas-cooled Reactors (HTGR) technology development and its modeling and simulation needs. There was a particular need for advanced modeling & simulation tools to predict thermal-fluid behavior in the nuclear reactor primary system, especially the core and lower and upper plena, during safety-related transients. In this report, two main such activities were presented: 1) system level HTGR simulations using system analysis code SAM for both normal operating and accident conditions; and 2) three-dimensional (3D) computational fluid dynamics (CFD) simulations on Deteriorated Turbulent Heat Transfer (DTHT) of interest to HTGR transients, including code benchmark and closure correlation development. In this work, Modular High Temperature Gas-cooled Reactor (MHTGR) full loop simulations were performed using System Analysis Module (SAM) code. Two approaches were used to model the reactor core; (i) unit-cell or lumped parameter model, and (ii) detailed ring model, to address the different needs under different reactor conditions. Aside from the core model changes other reactor components in the primary system, such as upper/lower plenum, heat exchanger and blower etc, were kept the same. In the unit-cell core model, the lumped parameter approach was used to model the coolant channels and the heat structures using 1-D and 2-D components, respectively. The fuel assembly heat structure was modeled based on unit-cell thickness approximation. In normal and controlled operating conditions, the calculated maximum temperatures for both coolant and heat structures were within the designed limits. The unit-cell model is, however, not able to capture the radial heat conduction through fuel assemblies, which is the dominant heat transfer mechanism during accident transients, such as pressurized conduction cooldown (PCC) event. To address this modeling challenge, ring models were developed. A simplified ring model was developed first. In this simplified model, the active core was simulated with 9 circular rings consists of 6 rings for homogenized fuel heat structure and 3 rings for gas coolant. It was quickly found that this model gives rather large simulation error as it was not able to correctly capture the solid structure representative geometries. A detailed ring model was developed to address this issue. The active core was simulated with 99 circular rings consists of 66 rings for homogenized fuel heat structure and 33 rings for gas coolant. In both ring models, six additional rings were included to represent inner reflector, outer reflector, core barrel, Reactor Pressure Vessel (RPV), RPV coolant channel, and Reactive Cavity Cooling System (RCCS). The detailed ring model was applied in two different SAM simulations; (i) normal operating conditions, and (ii) accident scenarios (more specifically, PCC event), for MHTGR design. In both simulations, the SAM calculated results for coolant and heat structure temperatures were within the reactor designed limits. The newly proposed detailed ring model approach in this work has proved to be a promising method to address the challenging modeling and simulation needs in the MHTGR design. As demonstrated in this work, this model was suitable for MHTGR simulations under both normal and accident conditions. Apart from the SAM model development for gas-cooled reactors, 3D Computational Fluid Dynamics (CFD) models were developed and validated with relaminarized Deteriorated Turbulent Heat Transfer (DTHT) MIT benchmark test. Two different CFD codes, Nek5000 and STAR-CCM+, were used for this benchmark activity. Compared with MIT data, Nek5000 shows very good agreement between its simulated and experimentally measured wall temperatures. In contrast, STAR-CCM+ under-predicted wall temperature due to over predicted turbulence in the wall region. The Nek5000 code was also used to develop wall-heat transfer correlations for laminar flow in cylindrical tube. Several simulations were performed for various Reynolds flow and wall heat fluxes. Significant deviations were found when compared the predicted heat transfer coefficient and the Sieder-Tate correlation. A new set of heat transfer correlation was proposed for laminar flow which is valid for 400<Re<2000.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Action-based dynamical modeling for the Milky Way disk

We present RoadMapping, a full-likelihood dynamical modeling machinery that aims to recover the Milky Way’s (MW) gravitational potential from large samples of stars in the Galactic disk. RoadMapping models the observed positions and velocities of stars with a parameterized, three-integral distribution function (DF) in a parameterized axisymmetric potential. We investigate through differential test cases with idealized mock data how the breakdown of model assumptions and data properties affect constraints on the potential and DF. Our key results are: (i) If the MW’s true potential is not included in the assumed model potential family, we can—in the axisymmetric case—still find a robust estimate for the potential, with only ≲10% difference in surface density within |z|⩽1.1 kpc inside the observed volume. (ii) Modest systematic differences between the true and model DF are inconsequential. E.g., when binning stars to define sub-populations with simple DFs, binning errors do not affect the modeling as long as the DF parameters of neighboring bins differ by <20%. In addition, RoadMapping ensures unbiased potential estimates for either (iii) small misjudgements of the spatial selection function (i.e., ≲15% at the survey volume’s edge), (iv) if distances are known to within 10%, or (v) if proper motion uncertainties are known within 10% or are smaller than δμ≲1 mas yr{sup −1}. Challenges are the rapidly increasing computational costs for large sample sizes. Overall, RoadMapping is well suited to making precise new measurements of the MW’s potential with data from the upcoming Gaia releases.

79 ASTRONOMY AND ASTROPHYSICS↗

U-net architected deep material network training with microstructure local field information

The Deep Material Network (DMN) has recently emerged as a powerful reduced-order modeling framework for simulating the mechanical response of heterogeneous materials such as composites. Unlike most data-driven approaches that directly learn a material’s response under prescribed loading, the DMN acts as a homogenization operator, learning the kinematic constraints and mechanical interactions of the underlying microstructure. However, traditional DMN training relies exclusively on homogenized effective properties derived from Direct Numerical Simulations (DNS), discarding the rich local field data that govern microstructural interactions. In this work, we extend the DMN framework to incorporate such local field information into the offline training process. Utilizing a U-Net architecture, we augment the DMN training objective to include the first and second statistical moments of the local stress fields obtained from linear DNS. This ensures that the learned network topology not only fits the effective stiffness but also accurately reflects the internal local stress and strain partitioning of the microstructure. The results confirm that supervising the localization process during training yields a superior surrogate model, reducing local prediction errors by an order of magnitude and significantly improving generalization to unseen nonlinear constitutive behaviors compared to traditional DMNs.

36 MATERIALS SCIENCE↗

Velocity-dependent J-factors for Milky Way dwarf spheroidal analogues in cosmological simulations

We study the impact of the dark matter velocity distribution modelling on signals from velocity-dependent dark matter annihilation in Milky Way dwarf spheroidal galaxies. Using the high resolution APOSTLE simulations, we identify analogues corresponding to Milky Way dwarf spheroidal galaxies, and from these directly determine the dark matter pair-wise relative velocity distribution, and compare to best-fitting Maxwell-Boltzmann distribution models. For three velocity-dependent annihilation models, p-wave, d-wave, and the Sommerfeld model, we quantify the errors introduced when using the Maxwell-Boltzmann parameterization. We extract a simple power-law relation between the maximum circular velocity of the dwarf spheroidal analogue and the peak speed of the Maxwell-Boltzmann distribution. We show that this relation can be used to accurately calculate the dark matter relative velocity distribution, and find that it allows us to estimate the dark matter annihilation signal without the need to directly calculate the relative velocity distribution for each galaxy. The scatter in the J-factors calculated from the analogues dominates the uncertainty obtained when compared to the J-factor as determined from the observational data for each dwarf spheroidal, with the largest scatter from d-wave models and the smallest from Sommerfeld models.

79 ASTRONOMY AND ASTROPHYSICS↗

ARC physics basis–magnetohydrodynamics

ARC is designed to produce 400 ⁢MW of net electricity and prove the commercial feasibility of a fusion power plant. In order to achieve this goal ARC has to operate with optimal core performance in a stationary scenario that minimises wear on the first wall and divertor. This requires avoiding or mitigating magnetohydrodynamic (MHD) instabilities which have the potential to not only degrade the plasma core but also lead to deleterious transient heat loads on plasma facing components. Therefore, this work aims at characterising the MHD stability of the high performance ARC scenario and inform the design of error field correction coils. Firstly, simulations of vertical displacement events show that an in-vessel coil is not needed and instead the poloidal shaping coils can be used to control vertical stability. These simulations also inform the demands on the corresponding coil power supplies. Stability analysis of the ideal kink mode with or without a conducting wall and kinetic effects suggests that the ARC baseline scenario operates deeply in the stable region. Using RDCON, tearing modes at the 𝑚/𝑛 =2/1 and 3/2 surfaces (with poloidal mode number 𝑚, and toroidal mode number 𝑛) are shown to be linearly stable, and including thermal transport effects in the rational surfaces lead to further stabilisation. However, other transient plasma instabilities can seed neoclassical tearing modes (NTMs). The marginally stable width of NTMs in ARC strongly depends on the internal inductance and can fall below 0.1% of the normalised poloidal flux. Furthermore, an empirical cross-machine model of the 𝑛 =1 error field leading to a disruption predicts a critical error field larger than SPARC but smaller than ITER. Three-dimensional coils can be designed with the Generalised Purturbed Equilbium Code based on a simple model that calculates the maximum correctable error field that is limited by the neoclassical toroidal viscosity torque. Broad scans of different coil geometries identify a set of 2 rows of off-midplane coils to be a suitable solution. It is also determined that such a set of three-dimensional coils is capable of correcting 𝑛 =2 error fields to some degree and creating strong enough 𝑛 =2 or 𝑛 =3 edge resonant perturbation fields for the suppression of edge-localised modes at reasonable coil currents. The final design of the first ARC will be further informed by results from SPARC.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

RBFNN-based Minimum Entropy Filtering for a Class of Stochastic Nonlinear Systems

This paper presents a novel minimum entropy filter design for a class of stochastic nonlinear systems which are subjected to non-Gaussian noises. Motivated by stochastic distribution control, an output entropy model is developed using RBF neural network while the parameters of the model can be identified by the collected data. Based upon the presented model, the filtering problem has been investigated while the system dynamics have been represented. As the model output is the entropy of the estimation error, the optimal nonlinear filter is obtained based on the Lyapunov design which makes the model output minimum. Moreover, the entropy assignment problem has been discussed as an extension of the presented approach. To verify the presented design procedure, a numerical example is given which illustrates the effectiveness of the presented algorithm. The contributions of this paper can be included as 1) an output entropy model is presented using neural network; 2) a nonlinear filter design algorithm is developed as the main result and 3) a solution of entropy assignment problem is obtained which is an extension of the presented framework.

42 ENGINEERING↗

Performance Efficient Multiresilience Using Checkpoint Recovery in Iterative Algorithms

In this paper, we address the design challenge of building multiresilient iterative high-performance computing (HPC) applications. Multiresilience in HPC applications is the ability to tolerate and maintain forward progress in the presence of both soft errors and process failures. We address the challenge by proposing performance models which are useful to design performance efficient and resilient iterative applications. The models consider the interaction between soft error and process failure resilience solutions. We experimented with a linear solver application with two distinct kinds of soft error detectors: one detector has high overhead and high accuracy, whereas the second has low overhead and low accuracy. We show how both can be leveraged for verifying the integrity of checkpointed state used to recover from both soft errors and process failures. Our results show the performance efficiency and resiliency benefit of employing the low overhead detector with high frequency within the checkpoint interval, so that timely soft error recovery can take place, resulting in less re-computed work.

Ashraf, Rizwan↗

Ligand Many-Body Expansion as a General Approach for Accelerating Transition Metal Complex Discovery

Methods that accelerate the evaluation of molecular properties are essential for chemical discovery. While some degree of ligand additivity has been established for transition metal complexes, it is underutilized in asymmetric complexes, such as the square pyramidal coordination geometries highly relevant to catalysis. To develop predictive methods beyond simple additivity, we apply a many-body expansion to octahedral and square pyramidal complexes and introduce a correction based on adjacent ligands (i.e., the cis interaction model). We first test the cis interaction model on adiabatic spin-splitting energies of octahedral Fe(II) complexes, predicting DFT-calculated values of unseen binary complexes to within an average of 1.4 kcal/mol. Uncertainty analysis reveals the optimal basis, comprising the homoleptic and mer symmetric complexes. We next show that the cis model (i.e., the cis interaction model solved for the optimal basis) infers both DFT- and CCSD(T)-calculated model catalytic reaction energies to within 1 kcal/mol on average. The cis model predicts low-symmetry complexes with reaction energies outside the range of binary complex reaction energies. We observe that trans interactions are unnecessary for most monodentate systems but can be important for some combinations of ligands, such as complexes containing a mixture of bidentate and monodentate ligands. Lastly, we demonstrate that the cis model may be combined with Δ-learning to predict CCSD(T) reaction energies from exhaustively calculated DFT reaction energies and the same fraction of CCSD(T) reaction energies needed for the cis model, achieving around 30% of the error from using the CCSD(T) reaction energies in the cis model alone.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Unraveling Excitonic Effects for the First Hyperpolarizabilities of Chromophore Aggregates

Excitonic interactions often significantly affect the optoelectronic properties of molecular materials. However, their role in determining the nonlinear optical response of organic electro-optic materials remains poorly understood. In this paper, we explore the effects of excitonic interactions on the first hyperpolarizability for aggregates of donor–acceptor chromophores. We show that calculations of the first hyperpolarizabilty of chromophore aggregates based on a two-state model agree well with the more rigorous coupled perturbed Hartree–Fock method. We then use both time-dependent density functional theory calculations and the molecular exciton approximation to parametrize the two-state model. Use of the molecular exciton approximation to the two-state model (i) is appropriate for disordered aggregates (unlike band theory), (ii) is computationally efficient enough for calculating the first hyperpolarizability of materials that consist of thousands of interacting chromophores, and (iii) allows the unraveling of the effects of both excitonic interactions and electrostatic polarization of the chromophore electron density by its environment on the first hyperpolarizability of molecular materials. We find that use of the molecular exciton approximation to the two-state model does not introduce significant additional errors compared to those introduced by applying the two-state model alone. We determine that the absolute change to the first hyperpolarizability of chromophore aggregates due to excitonic interactions increases with the size of the aggregate. For all sizes of disordered aggregates of chromophores considered in this paper, the inclusion of excitonic interactions on average decreases the magnitude of the first hyperpolarizability by 12–14% compared to the case of non-interacting chromophores. Lastly, we present a method for analytically calculating the first hyperpolarizability of a one-dimensional periodic array of chromophores within the molecular exciton approximation to the two-state model. This technique can be used to include an approximate correction for excitonic effects when simulating the electro-optic response of disordered and ordered organic materials.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Medium term municipal solid waste generation prediction by autoregressive integrated moving average

Generally, solid waste handling and management are performed by municipality or local authority. In most of developing countries, local authorities suffer from serious solid waste management (SWM) problems and insufficient data and strategic planning. Thus it is important to develop robust solid waste generation forecasting model. It helps to proper manage the generated solid waste and to develop future plan based on relatively accurate figures. In Malaysia, solid waste generation rate increases rapidly due to the population growth and new consumption trends that characterize the modern life style. This paper aims to develop monthly solid waste forecasting model using Autoregressive Integrated Moving Average (ARIMA), such model is applicable even though there is lack of data and will help the municipality properly establish the annual service plan. The results show that ARIMA (6,1,0) model predicts monthly municipal solid waste generation with root mean square error equals to 0.0952 and the model forecast residuals are within accepted 95% confident interval.

54 ENVIRONMENTAL SCIENCES↗

Calibr8 v.1.0

Calibr8 provides an application to rapidly prototype and perform material model calibration for complex plasticity models using advanced adjoint or forward sensitivity analyses for execution on massively parallel machines. These techniques can be orders of magnitude faster than traditional finite difference approaches for material model calibration. The underlying technology used in Calibr8 is automatic differentiation, which allows for the rapid implementation and testing of new plasticity models within its framework. Additionally, Calibr8 can perform adjoint-based error estimation to approximate discretization errors for user-implemented plasticity models. 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. SAND2021-10630 O

Granzow, BrianN.↗

SPIDARman: System-Level Physics-Informed Detection of Anomalies in Reactor Collected Data Considering Human Errors

In nuclear power plants (NPPs), anomalies arising from sensors or human errors (HEs) can undermine the performance and reliability of plant operations. Anomaly detection models can be employed to detect sensor errors and HEs. Additionally, physics-informed machine learning models can utilize the known physics of the system, as described by mathematical equations, to ensure that sensor values are consistent with physical laws. Hence, we propose SPIDARman: System-level Physics-Informed Detection of Anomalies in Reactor Collected Data Considering Human Errors, a holistic physics-informed anomaly detection approach based on generative adversarial networks (GANs) to detect anomalies in both automatically collected sensor data and manually collected surveillance data. Here we test our approach on data collected from a flow loop testbed, showcasing its potential to detect anomalies. Results demonstrate that the proposed model performs better than the baseline GAN-based models in detecting sensor and surveillance anomalies, suggesting the potential of physics-informed anomaly detection GAN models in NPPs.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Mitigating baryonic effects with a theoretical error covariance

ABSTRACT One of the primary sources of uncertainties in modelling the cosmic-shear power spectrum on small scales is the effect of baryonic physics. Accurate cosmology for stage-IV surveys requires knowledge of the matter power spectrum deep in the non-linear regime at the per cent level. Therefore, it is important to develop reliable mitigation techniques to take into account baryonic uncertainties if information from small scales is to be considered in the cosmological analysis. In this work, we develop a new mitigation method for dealing with baryonic physics for the case of the shear angular power spectrum. The method is based on an augmented covariance matrix that incorporates baryonic uncertainties informed by hydrodynamical simulations. We use the results from 13 hydrodynamical simulations and the residual errors arising from a fit to a ΛCDM model using the extended halo model code HMCode to account for baryonic physics. These residual errors are used to model a so-called theoretical error covariance matrix that is added to the original covariance matrix. In order to assess the performance of the method, we use the 2D tomographic shear from four hydrodynamical simulations that have different extremes of baryonic parameters as mock data and run a likelihood analysis comparing the residual bias on Ωm and σ8 of our method and the HMCode for an LSST-like survey. We use different modelling of the theoretical error covariance matrix to test the robustness of the method. We show that it is possible to reduce the bias in the determination of the tested cosmological parameters at the price of a modest decrease in the precision.

79 ASTRONOMY AND ASTROPHYSICS↗

Optimized Sampling Strategies For Non-Proliferation Monitoring: Report

Concentration data collected from the 2013 H-Canyon effluent reprocessing experiment were reanalyzed to improve the source term estimate. When errors in the model-predicted wind speed and direction were removed, the source term uncertainty was reduced to 30% of the mean. This explained the factor of 30 difference between the source term size derived from data at 5 km and 10 km downwind in terms of the time history of dissolution. The results show a path forward to develop a sampling strategy for quantitative source term calculation.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Non-local thermodynamic equilibrium 1.5D modeling of red giant stars

Spectra for two-dimensional (2D) stars in the 1.5D approximation are created from synthetic spectra of one-dimensional (1D) non-local thermodynamic equilibrium (NLTE) spherical model atmospheres produced by the PHOENIX code. The 1.5D stars have the spatially averaged Rayleigh-Jeans flux of a K3-4 III star while varying the temperature difference between the two 1D component models (ΔT {sub 1.5D}) and the relative surface area covered. Synthetic observable quantities from the 1.5D stars are fitted with quantities from NLTE and local thermodynamic equilibrium (LTE) 1D models to assess the errors in inferred T {sub eff} values from assuming horizontal homogeneity and LTE. Five different quantities are fit to determine the T {sub eff} of the 1.5D stars: UBVRI photometric colors, absolute surface flux spectral energy distributions (SEDs), relative SEDs, continuum normalized spectra, and TiO band profiles. In all cases except the TiO band profiles, the inferred T {sub eff} value increases with increasing ΔT {sub 1.5D}. In all cases, the inferred T {sub eff} value from fitting 1D LTE quantities is higher than from fitting 1D NLTE quantities and is approximately constant as a function of ΔT {sub 1.5D} within each case. The difference between LTE and NLTE for the TiO bands is caused indirectly by the NLTE temperature structure of the upper atmosphere, as the bands are computed in LTE. We conclude that the difference between T {sub eff} values derived from NLTE and LTE modeling is relatively insensitive to the degree of the horizontal inhomogeneity of the star being modeled and largely depends on the observable quantity being fit.

79 ASTRONOMY AND ASTROPHYSICS↗

Measurement-induced purification in large- N hybrid Brownian circuits

Competition between unitary dynamics that scrambles quantum information nonlocally and local measurements that probe and collapse the quantum state can result in a measurement-induced entanglement phase transition. Here we study this phenomenon in an analytically tractable all-to-all Brownian hybrid circuit model composed of qubits. The system is initially entangled with an equal sized reference, and the subsequent hybrid system dynamics either partially preserves or totally destroys this entanglement depending on the measurement rate. Our approach can access a variety of entropic observables which are distinguished by the averaging procedure, and for concreteness we focus on a particular purity quantity for which the averaging is particularly simple. We represent the purity as a path integral coupling four replicas with twisted boundary conditions. Saddle-point analysis reveals a second-order phase transition corresponding to replica permutation symmetry breaking below a critical measurement rate. The transition is mean-field-like and we characterize the critical properties near the transition in terms of a simple Ising field theory in 0 + 1 dimensions. In addition to studying the purity of the entire system, we study subsystem purities and relate these results to manifestations of quantum error correction in the model. Here, we also comment on the experimental feasibility for simulating this averaged purity, and corroborate our results with exact diagonalization for modest system sizes.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Validation of the High-Resolution Salish Sea Tidal Hydrodynamic Model

In this study, a tidal hydrodynamic model was developed and validated to simulate tidal currents in Puget Sound, Washington, to support tidal energy resource characterization using the unstructured-grid, Finite Volume Community Ocean Model (FVCOM). The Salish Sea tidal hydrodynamic model was driven by tides along two open boundaries at the entrance of the Strait of Juan de Fuca and north end of Georgia Strait, and river flows from 19 major rivers in the Salish Sea. To simulate the tidal current in Puget Sound, a high-resolution model grid is required to accurately represent the complex coastlines and bathymetry. The spatial resolution of the model grid varies from ~10 m near river boundaries and ~30 m in small tidal channels and estuaries to near 1000 m inside Georgia Strait and at the open boundaries. Model validation was carried out by comparing simulated and observed water levels at 12 tidal stations and currents at 135 Acoustic Doppler Current Profiler stations in the model domain. A set of model performance metrics, including root mean square error, scatter index, bias, and linear correlation coefficient, were used to quantify the model skills in simulating the tidal hydrodynamics in Puget Sound. Error statistics showed an overall good agreement between simulated and observed tidal elevations and currents, which demonstrated that the Puget Sound tidal model can be used to accurately characterize the tidal stream energy resource in Puget Sound.

16 TIDAL AND WAVE POWER↗

A performance model for GPUs with caches

To exploit the abundant computational power of the world's fastest supercomputers, an even workload distribution to the typically heterogeneous compute devices is necessary. While relatively accurate performance models exist for conventional CPUs, accurate performance estimation models for modern GPUs do not exist. This paper presents two accurate models for modern GPUs: a sampling-based linear model, and a model based on machine-learning (ML) techniques which improves the accuracy of the linear model and is applicable to modern GPUs with and without caches. We first construct the sampling-based linear model to predict the runtime of an arbitrary OpenCL kernel. Based on an analysis of NVIDIA GPUs' scheduling policies we determine the earliest sampling points that allow an accurate estimation. The linear model cannot capture well the significant effects that memory coalescing or caching as implemented in modern GPUs have on performance. We therefore propose a model based on ML techniques that takes several compiler-generated statistics about the kernel as well as the GPU's hardware performance counters as additional inputs to obtain a more accurate runtime performance estimation for modern GPUs. We demonstrate the effectiveness and broad applicability of the model by applying it to three different NVIDIA GPU architectures and one AMD GPU architecture. On an extensive set of OpenCL benchmarks, on average, the proposed model estimates the runtime performance with less than 7 percent error for a second-generation GTX 280 with no on-chip caches and less than 5 percent for the Fermi-based GTX 580 with hardware caches. On the Kepler-based GTX 680, the linear model has an error of less than 10 percent. On an AMD GPU architecture, Radeon HD 6970, the model estimates with 8 percent of error rates. As a result, the proposed technique outperforms existing models by a factor of 5 to 6 in terms of accuracy.

97 MATHEMATICS AND COMPUTING↗