Search NASA⌕ Search

SEARCH · Search NASA

Results for “production simulation codes”

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 1,099 records · Page 61

Development of a BlueCRAB/MELCOR Framework for Supporting Realistic Mechanistic Source Term Calculations in Microreactors

Efforts are currently underway to deploy microreactor modeling and simulation tools to better support vendors and regulatory authorities in submitting and reviewing licensing applications. In particular, the Nuclear Regulatory Commission is expected to rely on the Comprehensive Reactor Analysis Bundle (BlueCRAB) multiphysics toolset in performing design- and beyond-design-basis accident analyses. In addition, the Nuclear Regulatory Commission has been using the MELCOR code to estimate mechanistic source terms during accidents. As MELCOR relies on isotopic inventory and reactor temperature/power evolution profiles during accident conditions—all of which can theoretically be obtained from BlueCRAB—the ultimate goal of this activity is to establish a common BlueCRAB-MELCOR framework. However, prior to the present research, BlueCRAB had never been used to calculate such quantities of interest at the full-core level. While there are many Monte Carlo (MC) codes capable of computing such quantities of interest, they are unable to readily account for multiphysics feedback. BlueCRAB allows for the coupling of different physics codes together to perform multiphysics-informed calculations. Therefore, the purpose of this fiscal year 2023 work is to investigate the feasibility and challenges of performing such calculations within BlueCRAB so as to generate the data that MELCOR relies on. To demonstrate the methodology, the proposed workflow was applied to a prototypical heat pipe-cooled microreactor model. To predict isotopic concentrations (taking into account the ac- cumulation of fission products during operation), the necessary microscopic cross sections were generated via OpenMC and tabulated with respect to temperature and burnup. Next, a recently developed capability in Griffin (the reactor physics application in BlueCRAB) was used to convert the OpenMC output format into the ISOXML format used by Griffin. A multiphysics microscopic depletion calculation that involved performing a coupled full-core, heterogeneous neutron trans- port and thermal calculation at each depletion step was conducted to deplete the core to end of life (EOL) conditions so as to provide both isotopics and the initial condition for the transient calculation. Following a brief null-transient to verify that the initial condition had been properly restarted and was indeed in thermal equilibrium, a heat pipe failure transient was simulated. Thus, the entire workflow of using BlueCRAB to generate MELCOR inputs, from cross-section generation to producing isotopic inventory and power/temperature evolution profiles during transients, is demonstrated. This report also details the identified gaps in the workflow and how they were (for the most part) addressed. Future work should focus on directly including MEL- COR into the workflow by performing a MELCOR calculation using the BlueCRAB-generated input data. In addition, the heat pipe reactor design should be improved so as to reflect more prototypical burnup characteristics at EOL.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

SU-F-T-656: Monte Carlo Study On Air Activation Around a Medical Electron Linac

Purpose: In high energy photon therapy, several radiation protection issues result from photonuclear reactions. The activation of air - directly by photonuclear reactions as well as indirectly by capture of photoneutrons generated inside the linac head - is a major point of concern for the medical staff. The purpose of this study was to estimate the annual effective dose to medical workers due to activated air around a medical high energy electron linac by means of Monte Carlo simulations. Methods: The treatment head of a Varian Clinac in 18 MV-X mode as well as the surrounding concrete bunker were modeled and the radiation transport was simulated using the Monte Carlo code FLUKA, starting from the primary electron striking the bremsstrahlung target. The activation yields in air from photo-disintegration of O-16 and N-14 nuclei as well as from neutron capture on Ar-40 nuclei were obtained from the simulations. The activation build-up, radioactive decay and air ventilation were studied using a mathematical model. The annual effective dose to workers was estimated by using published isotope specific conversion factors. Results: The oxygen and nitrogen activation yields were in contrast to the argon activation yield found to be field size dependent. The impact of the treatment room ventilation on the different air activation products was investigated and quantified. An estimate with very conservative assumptions gave an annual effective dose to workers of < 1 mSv/a. Conclusion: From the results of this study it can be concluded that the contribution of air activation to the radiation exposure to medical workers should be negligible in modern photon therapy, especially when it is compared to the dose due to prompt neutrons and the activation of heavy solid materials such as the jaws and the collimators inside the linac head.

60 APPLIED LIFE SCIENCES↗

Effects of delta ray electrons on measurement uncertainties of harp system

A harp system, which is a multi-wire beam profile monitoring (MWPM) system, is planned upstream of the spallation target to make in situ calibration of beam current density configuration on the target along with beam imaging from luminescent coating on the beam entrance window at the Second Target Station (STS) of the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (ORNL). This beam interception-based beam diagnostics system on the target will be used to ensure that the maximum beam loads on the target are within the design range during neutron production. Current design of the harp consists of three layers of measurement wires each of which is sandwiched between voltage biasing wire planes. The signal obtained from each measurement wire layer is disturbed by secondary electrons (SE) and delta rays produced by beam-matter interactions in neighboring wires and ionization of residual gases in accelerator vacuum. While the backgrounds from SE can be suppressed by voltage biasing, the delta-ray electrons with kinetic energies above keV ranges overcome the electric potential bias. In this paper, we study the effects of delta-rays on the measurement uncertainties of MWPM using the particle transport simulation code FLUKA. Furthermore, the cases where the harp system is installed in the proximity of a large delta ray sources such as proton beam window or in the core vessel filled with sub-atmospheric gas have been studied.

Lee, Yong Joong↗

Combining Ratio Estimation for Low Density Parity Check (LDPC) Coding

The Low Density Parity Check (LDPC) Code decoding algorithm make use of a scaled receive signal derived from maximizing the log-likelihood ratio of the received signal. The scaling factor (often called the combining ratio) in an AWGN channel is a ratio between signal amplitude and noise variance. Accurately estimating this ratio has shown as much as 0.6 dB decoding performance gain. This presentation briefly describes three methods for estimating the combining ratio: a Pilot-Guided estimation method, a Blind estimation method, and a Simulation-Based Look-Up table. The Pilot Guided Estimation method has shown that the maximum likelihood estimates of signal amplitude is the mean inner product of the received sequence and the known sequence, the attached synchronization marker (ASM) , and signal variance is the difference of the mean of the squared received sequence and the square of the signal amplitude. This method has the advantage of simplicity at the expense of latency since several frames worth of ASMs. The Blind estimation method s maximum likelihood estimator is the average of the product of the received signal with the hyperbolic tangent of the product combining ratio and the received signal. The root of this equation can be determined by an iterative binary search between 0 and 1 after normalizing the received sequence. This method has the benefit of requiring one frame of data to estimate the combining ratio which is good for faster changing channels compared to the previous method, however it is computationally expensive. The final method uses a look-up table based on prior simulated results to determine signal amplitude and noise variance. In this method the received mean signal strength is controlled to a constant soft decision value. The magnitude of the deviation is averaged over a predetermined number of samples. This value is referenced in a look up table to determine the combining ratio that prior simulation associated with the average magnitude of the deviation. This method is more complicated than the Pilot-Guided Method due to the gain control circuitry, but does not have the real-time computation complexity of the Blind Estimation method. Each of these methods can be used to provide an accurate estimation of the combining ratio, and the final selection of the estimation method depends on other design constraints.

Mahmoud, Saad↗

Combined Experimental and Computational Approach to Predict the Glass-Water Reaction

The use of mineral and glass dissolution rates measured in laboratory experiments to predict the weathering of primary minerals and volcanic and nuclear waste glasses in field studies requires the construction of rate models that accurately describe the weathering process over geologic timescales. Additionally, the need to model the long-term behavior of nuclear waste glass for the purpose of estimating radionuclide release rates requires that rate models be validated with long-term experiments. Several long-term test methods have been developed to accelerate the glass-water reaction [drip test, vapor hydration test, product consistency test B, and pressurized unsaturated flow (PUF)], thereby reducing the duration required to evaluate long-term performance. Currently, the PUF test is the only method that mimics the unsaturated hydraulic properties expected in a subsurface disposal facility and simultaneously monitors the glass-water reaction. PUF tests are being conducted to accelerate the weathering of glass and validate the model parameters being used to predict long-term glass behavior. A one-dimensional reactive chemical transport simulation of glass dissolution and secondary phase formation during a 1.5-year-long PUF experiment was conducted with the Subsurface Transport Over Reactive Multiphases (STORM) code. Results show that parameterization of the computer model by combining direct bench scale laboratory measurements and thermodynamic data provides an integrated approach to predicting glass behavior over the length of the experiment. Over the 1.5-year-long test duration, the rate decreased from 0.2 to 0.01 g/(m2 day) based on B release for low-activity waste glass LAWA44. The observed decrease is approximately two orders of magnitude higher than the decrease observed under static conditions with the SON68 glass (estimated to be a decrease by four orders of magnitude) and suggests that the gel-layer properties are less protective under these dynamic conditions.

aluminoborosilicate nuclear waste glass↗

Criticality and axial offset searches based on the integrated neutron balance approach

Criticality and axial offset (AO) searches are key features in the simulation of plant operation procedures. Typically, the dedicated algorithms consist of either a Newton method or a secant method that features the successive computation of difference quotients to be used as derivatives. Within this context, computational robustness and efficiency are of paramount importance. This applies in particular to when the variations imposed during the search are spatially heterogeneous, such as in the case of tuning control rod positions for achieving target AO. In order to optimize this kind of search capability in AREVA NP's reactor code ARTEMIS in accordance with this design principle, a generalized approach has been implemented that harmonizes very well numerically with the overall multi-physics iterative solution process. As embedded in the latter, the new neutronic integral re-balancing approach is defined through periodic whole-core space-energy integrations of the principal terms in the core-integrated process-rate balance terms (i.e. neutron absorption, production and leakage). This procedure yields sequences of single zero-dimensional equations from which the chosen tuning parameters can be solved directly in dependence of the imposed (and systematically fixed) values for the response quantities k{sub eff} and/or AO. The converged result of the iteration sequence of successively obtained search parameter values is final in terms of being fully consistent with the entire set of multi-physics equations while enabling the accurate fulfillment of the target response value. The k etc and AO searches can be pursued simultaneously. Judging from the results of pursued verifications, the neutronic integral re-balancing approach fulfils the above-mentioned expectations convincingly. Specific verification examples are presented, such as the determination of the insertion depth of a critical control bank, a double search on target criticality and target AO by adjusting the boron concentration and the position of a control bank, and a critical power search during reactor cycle stretch out. (authors)

97 MATHEMATICS AND COMPUTING↗

Cluster dynamics simulation of xenon diffusion during irradiation in UO 2

Diffusion of fission gas in UO 2 nuclear fuel impacts several important performance metrics, such as fission gas release, swelling, and thermal conductivity. Current empirical models of fission gas release have significant uncertainty, some of which derives from the bulk diffusion rate and its dependence on, for example, fuel chemistry and irradiation. In this work, we have applied the previously-developed Free Energy Cluster Dynamics (FECD) methodology in the code Centipede to calculate xenon cluster concentrations in UO 2 under intrinsic (high temperature) and irradiation-enhanced (intermediate temperature) conditions in order to develop a model of the xenon diffusion coefficient based on the atomic scale mechanisms responsible for transport. While the diffusion mechanism for xenon in UO 2 is adequately described by the Xe + U 2 O vacancy cluster for intrinsic conditions, a similar process is not capable of capturing measured in-pile fission gas diffusivity at intermediate temperatures. Therefore, a different diffusion mechanism must dominate under this regime. Using calculated atomistic data, we have shown that irradiation-enhanced diffusion at intermediate temperatures occurs via the larger Xe + U 4 O y vacancy clusters, which have lower migration barriers and increase in concentration by several orders of magnitude compared to intrinsic conditions. This mechanism is enabled by the increased uranium vacancy concentration under irradiation due to Frenkel pair production. In addition, the fast migration of uranium interstitials with two attached oxygen interstitials lowers the total uranium interstitial concentration through reactions with sinks. This allows the extended defects, such as Xe + U 4 O y vacancy clusters, to maintain high concentrations by limiting annihilation with attached vacancies. Furthermore, predictions using the Xe + U 4 O y diffusion mechanism are in good agreement with experiment, albeit with some differences in the Arrhenius slope, which we believe may be related to either experimental or model parameter uncertainty. Lastly, an analytical expression suitable for application in fuel performance simulations was derived to capture the predictions of the Centipede simulations.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Loss Factor of Tapered Structures for Short Bunches

Using the electromagnetic simulation code ECHO, we have found a simple phenomenological formula that accurately describes the loss factor for short bunches traversing an axisymmetric tapered collimator. In this paper, we consider tapered collimators with rectangular cross-section and use the GdfidL code to calculate the loss factor dependence on the geometric parameters for short bunches. The results for both axisymmetric and rectangular collimators are discussed. The behaviour of the impedance of tapered structures for very short bunches in the optical regime has been determined in refs. [10,11]. Here, for the loss factors for two particular geometries, we have studied the departure from the optical regime behaviour as bunch length is increased. In both cases, the ratio of the loss factor for the tapered collimator to the loss factor in the optical regime is a function only of the scaling parameter {sigma}L/d{sup 2}. The fact that the bunch length a and the taper length L appear as a product is consistent with the recent scaling derived by Stupakov in ref. [12], since there is only a weak dependence on g. One noteworthy fact that is not a priori expected is that only the larger radius or vertical half-aperture d appears. The reduction factor is independent of b. Moreover, it is striking that the specific form involving the arctan given in Eq. (5) holds for both geometries, with only the coefficient {mu} differing by a factor of {approx}2 for flat vs round. This suggests that there may be a useful phenomenological form for more general geometries which may follow from natural extensions of Eq. (5). This possibility is presently being investigated.

43 PARTICLE ACCELERATORS↗

Structure of oxides prepared by decomposition of layered double Mg–Al and Ni–Al hydroxides

Abstracts: Thermal decomposition of Mg–Al and Ni–Al layered double hydroxides LDH at temperatures lower than 800 °C leads to the formation of oxides with different structures. Mg–Al oxide has a very defective structure and consists of octahedral layers as in periclase MgO and mixed octahedral–tetrahedral layers as in spinel MgAl{sub 2}O{sub 4}. Mixed Ni–Al oxide has a sandwich-like structure, consisting of a core with Al-doped NiO-like structure and some surface layers with spinel NiAl{sub 2}O{sub 4} structure epitaxial connected with the core. Suggested models were verified by simulation of X-ray diffraction patterns using DIFFaX code, as well as HRTEM, IR-, UV-spectroscopies, and XPS. - Graphical abstract: In the Mg–Al layered double hydroxide Al{sup 3+} ions migrate into interlayers during decomposition. The Mg–Al oxide represents sequence of octahedral and octahedral–tetrahedral spinel layers with vacancies. The Ni–Al oxide has a sandwich-like structure with NiO-like core and surface spinel layers as a result of migration of Al{sup 3+} ions on the surface. The models explain the presence and absence of “memory effect” for the Mg–Al and Ni–Al oxides, respectively. - Highlights: • We study products of Mg(Ni)–Al LDH decomposition by calcination at 500(400)–800 °C. • In Mg–Al/Ni–Al LDH Al ions migrate into interlayers/on the surface during decomposition. • Mg–Al oxide represents sequence of periclase- and spinel-like layers with vacancies. • Ni–Al oxide has a sandwich-like structure with NiO-like core and surface spinel layers. • The models explain the presence/absence of “memory effect” for Mg–Al/Ni–Al oxides.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

NEML2 - THE NEW ENGINEERING MATERIAL MODEL LIBRARY, VERSION 2

SF-22-157 NEML2 is an extension of an earlier open source code developed at ANL (SF-xxxx) called NEML. These libraries provide tools for constructing and using models of engineering materials, with a focus on high temperature materials for nuclear reactor applications. The software include tools for setting up material models, calibrating models, storing models for later use in other simulations, and running simple "material point" simulations of the material behavior. The main difference between NEML and NEML2 is the new software has been complete restructured to run on GPUs and other accelerators in order to improve the performance of methods used to train constitutive models against data and to improve the performance of the final models in production environments, like finite element software. While over time we expect NEML2 to become more-or-less feature complete with NEML we are filing a new software disclosure because (1) the library has been complete restructured and (2) we want to host NEML2 in a separate repository.

MESSNER, MARK↗

An expanding vortex site for the r-process in rotating stellar collapse

The astrophysical site of r-process nucleosynthesis is investigated theoretically using two-dimensional expanding-vortex stellar-collapse simulations based on the Eulerian adaptive-mesh MHD code of Symbalisty (1984). The results are presented graphically, and it is found that the classical r-process can be explained as the result of the collapse of the highly rotating iron core of a 15-solar-mass star, in which the angular velocity along the rotation axis reaches a maximum of 1000 rad/s, corresponding to an angular momentum of 3.4 x 10 to the 48th erg/s for the inner 1.5 solar mass. The ejected jets are shown to yield about 0.0004 solar mass per supernova, sufficient to explain the observed abundances of r-process products.

Symbalisty, E. M. D.↗

Numerical simulation of hypersonic inlet flows with equilibrium or finite rate chemistry

An efficient numerical program incorporated with comprehensive high temperature gas property models has been developed to simulate hypersonic inlet flows. The computer program employs an implicit lower-upper time marching scheme to solve the two-dimensional Navier-Stokes equations with variable thermodynamic and transport properties. Both finite-rate and local-equilibrium approaches are adopted in the chemical reaction model for dissociation and ionization of the inlet air. In the finite rate approach, eleven species equations coupled with fluid dynamic equations are solved simultaneously. In the local-equilibrium approach, instead of solving species equations, an efficient chemical equilibrium package has been developed and incorporated into the flow code to obtain chemical compositions directly. Gas properties for the reaction products species are calculated by methods of statistical mechanics and fit to a polynomial form for C(p). In the present study, since the chemical reaction time is comparable to the flow residence time, the local-equilibrium model underpredicts the temperature in the shock layer. Significant differences of predicted chemical compositions in shock layer between finite rate and local-equilibrium approaches have been observed.

Yu, Sheng-Tao↗

Contributions of TetrUSS to Project Orion

The NASA Constellation program has relied heavily on Computational Fluid Dynamics simulations for generating aerodynamic databases and design loads. The Orion Project focuses on the Orion Crew Module and the Orion Launch Abort Vehicle. NASA TetrUSS codes (GridTool/VGRID/USM3D) have been applied in a supporting role to the Crew Exploration Vehicle Aerosciences Project for investigating various aerodynamic sensitivities and supplementing the aerodynamic database. This paper provides an overview of the contributions from the TetrUSS team to the Project Orion Crew Module and Launch Abort Vehicle aerodynamics, along with selected examples to highlight the challenges encountered along the way. A brief description of geometries and tasks will be discussed followed by a description of the flow solution process that produced production level computational solutions. Four tasks conducted by the USM3D team will be discussed to show how USM3D provided aerodynamic data for inclusion in the Orion aero-database, contributed data for the build-up of aerodynamic uncertainties for the aero-database, and provided insight into the flow features about the Crew Module and the Launch Abort Vehicle.

Mcmillin, Susan N.↗

Combined Experimental and Computational Approach to Predict the Glass-Water Reaction

The use of mineral and glass dissolution rates measured in laboratory experiments to predict the weathering of primary minerals and volcanic and nuclear waste glasses in field studies requires the construction of rate models that accurately describe the weathering process over geologic time-scales. Additionally, the need to model the long-term behavior of nuclear waste glass for the purpose of estimating radionuclide release rates requires that rate models are validated with long-term experiments. Several long-term test methods have been developed to accelerate the glass-water reaction [drip test, vapor hydration test, product consistency test-B, and pressurized unsaturated flow (PUF)], thereby reducing the duration required to evaluate long-term performance. Currently, the PUF test is the only method that mimics the unsaturated hydraulic properties expected in a subsurface disposal facility and simultaneously monitors the glass-water reaction. PUF tests are being conducted to accelerate the weathering of glass and validate the model parameters being used to predict long-term glass behavior. A one-dimensional reactive chemical transport simulation of glass dissolution and secondary phase formation during a 1.5-year long PUF experiment was conducted with the subsurface transport over reactive multi-phases code. Results show that parameterization of the computer model by combining direct bench-scale laboratory measurements and thermodynamic data provides an integrated approach to predicting glass behavior over the length of the experiment. Over the 1.5-year long test duration, the rate decreased from 0.2 to 0.01 g/(m2 d) base on B release. The observed decrease is approximately two orders of magnitude higher than the decrease observed under static conditions with the SON68 glass (estimated to be a decrease by 4 orders of magnitude) and suggest the gel-layer properties are less protective under these dynamic conditions.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

JASMIN: Japanese-American study of muon interactions and neutron detection

Experimental studies of shielding and radiation effects at Fermi National Accelerator Laboratory (FNAL) have been carried out under collaboration between FNAL and Japan, aiming at benchmarking of simulation codes and study of irradiation effects for upgrade and design of new high-energy accelerator facilities. The purposes of this collaboration are (1) acquisition of shielding data in a proton beam energy domain above 100GeV; (2) further evaluation of predictive accuracy of the PHITS and MARS codes; (3) modification of physics models and data in these codes if needed; (4) establishment of irradiation field for radiation effect tests; and (5) development of a code module for improved description of radiation effects. A series of experiments has been performed at the Pbar target station and NuMI facility, using irradiation of targets with 120 GeV protons for antiproton and neutrino production, as well as the M-test beam line (M-test) for measuring nuclear data and detector responses. Various nuclear and shielding data have been measured by activation methods with chemical separation techniques as well as by other detectors such as a Bonner ball counter. Analyses with the experimental data are in progress for benchmarking the PHITS and MARS15 codes. In this presentation recent activities and results are reviewed.

43 PARTICLE ACCELERATORS↗

Experimental characterization and potential energy savings of insulated cladding for U.S. residential buildings

This study evaluates a novel insulated cladding composed of foamed cement and vinyl siding that is designed as a retrofit product to enhance thermal performance in U.S. residential buildings. Experimental measurements showed that foamed cement exhibits a thermal conductivity of 0.0334–0.0365 W/m-K (R-value of 4.0–4.3 °F⋅ft 2 ⋅h/BTU), which is significantly higher than that of conventional fiber cement boards. Although mechanical testing confirmed the material’s brittleness and low strength, it also indicated its suitability for nonstructural insulation applications. Water vapor permeability testing demonstrated effective moisture resistance with the polymer coating. To evaluate the insulated cladding’s energy performance, we used the U.S. Department of Energy’s prototype single-family building model and simulated across 50 locations spanning 16 International Energy Conservation Code climate zones. The R-7 (R-value of 7 °F⋅ft 2 ⋅h/BTU per inch) retrofit cladding resulted in heating energy savings up to 275 therms (Fairbanks, AK) and cooling energy savings up to 1,547 kWh (Phoenix, AZ). The energy cost savings varied by region, but the highest annual cost savings ($\$$434/year) were observed in Santa Maria, CA, and the highest annual cost savings percentage (37 %) were noted in Monterey, CA. On a national scale, the insulated retrofit provided an average annual energy cost reduction of 16 %, highlighting the cladding system’s broad applicability and financial viability.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Dissipation and particle energization in moderate to low beta turbulent plasma via PIC simulations

Here, we simulate decaying turbulence in electron-positron pair plasmas using a fully-kinetic particle-in-cell (PIC) code. We run two simulations with moderate-to-low plasma β (the ratio of thermal pressure to magnetic pressure). The energy decay rate is found to be similar in both cases. The perpendicular wave-number spectrum of magnetic energy shows a slope between ${k}_{\perp }^{-1.3}$ and ${k}_{\perp }^{-1.1}$, where the perpendicular (⊥) and parallel (∥) directions are defined with respect to the magnetic field. The particle kinetic energy distribution function shows the formation of a non-thermal feature in the case of lower plasma β, with a slope close to E-1. The correlation between thin turbulent current sheets and Ohmic heating by the dot product of electric field (E) and current density (J) is investigated. By heating the parallel E∥ centerdot J∥ term dominates the perpendicular E⊥ centerdot J⊥ term. Regions of strong E∥ centerdot J∥ are spatially well-correlated with regions of intense current sheets, which also appear correlated with regions of strong E∥ in the low β simulation, suggesting an important role of magnetic reconnection in the dissipation of low β plasma turbulence.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Evaluating the effectiveness of program data features for guiding memory management

Recent trends have led to the adoption of larger and more complex memory systems, often with multiple tiers of memory performance within the same platform. To utilize complex memory systems efficiently, current data management strategies must be altered to map usage demands to the underlying hardware. Applications, as the generators of memory accesses, are well-suited to guide this process, but building and maintaining separate source code versions for different memory systems is not feasible in most cases. One potential solution is to employ automated program profiling and analysis to facilitate the production of application-based guidance. By attaching memory usage information to static or lightweight program features, compilers and runtime systems can generate fine-grained guidance without additional efforts from users or developers. Recent works have employed this approach with some success, but it is not clear which program features are most useful for guiding data management.This work evaluates the effectiveness of using different program data features to predict memory usage and guide memory management. It employs a custom set of simulation tools, based in the Intel® Pin framework, to collect and analyze the usage characteristics of application data associated with common program data features, such as allocation sites, types, and context. The results show that even relatively simple features, such as object size, are useful for selecting data with similar usage properties, but finer-grained features, such as the instructions that access a particular object, are often much more effective. Additionally, this work evaluates the performance of using different data features and different program inputs to guide data placement on a heterogeneous memory platform with a limited amount of high performance memory.

Effler, T. Chad↗