Search NASA⌕ Search

SEARCH · Search NASA

Results for “computer system benchmarking”

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 775 records · Page 43

FY23 Progress on Computational Modeling of the Water-Based NSTF

This report summarizes the system-level modeling effort by Argonne National Laboratory (Argonne) of the Natural convection Shutdown heat removal Test Facility (NSTF) in FY23. As a continuation of the modeling effort from FY22, this year’s work focuses on improving the RELAP5-3D model developed previously for two-phase flow simulations. The RELAP5-3D model is updated to more accurately capture the heat loss experienced by the facility. The updated model is compared against experimental data for benchmarking purposes of the RELAP5-3D input model. By correctly accounting for heat loss, the updated RELAP5-3D model can now predict the two-phase baseline case more accurately. The onset and the duration of instability are captured well by the model. Furthermore, analyses are performed to better understand the instability mechanism experienced by the flow where the expansion of the boiling boundary in the chimney is studied in details and the fundamental frequencies of the oscillations are obtained. The updated RELAP5-3D model is further compared against four fault conditions, namely the reduction of riser header inlet flow area, depletion of system inventory, blocked riser channels, and static boiling scenario. For each fault condition, minor modifications and tuning are performed to improve the predictions of the model. The purpose of the analyses is to investigate the capability of RELAP5-3D in predicting complex two-phase flows in possible accident scenarios in actual Reactor Cavity Cooling System (RCCS). Overall, the model is able to capture the behaviors and trends of these fault conditions relatively well. Some discrepancies remain between the experimental data and the predictions, many of which are likely due to the differences in the predicted and experimental vapor generation rate. Future work will focus on the continued development of the current RELAP5-3D input model of the NSTF to both improve the accuracy of the model’s predictive capability and continue supporting the experimental program needs. The mutually beneficial relationship between analysis and experimental efforts has become integral to the parent NSTF program, and the greater objective to fully understand and accurately predict the heat removal performance of a full scale RCCS concept.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

A time-parallel method for scalable heat transfer simulations of additive manufacturing

Here, a major challenge in simulating the thermal behavior in additive manufacturing processes is the disparate length and time scales between transport phenomena occurring in the melt pool and the component. A common simulation approach relies on spatial decomposition for parallel computing, but due to the nature of heat transfer in AM, where most of the computational expenditure is localized near the melt pool, the computational speedup from spatial parallelization saturates quickly. Therefore, additional parallelism by means of time-domain decomposition is needed to fully take advantage of high-performance computing (HPC) resources. This work introduces a time-parallel method to improve the computational scalability of additive manufacturing simulations on HPC systems, while maintaining high temporal resolution of heat transfer near the melt pool. The method, inspired by the nonlinear paraexp formalism, performs an iterative superposition of nonlinear solutions to the initial value problem, integrating the heat equation across overlapping time-parallel intervals. For a single layer of the NIST AMB2018–01 L7 benchmark problem, the method achieves a 38.51x speedup in wall-clock time with a maximum error in the global temperature solution of 0.99%. This reduces the total solution time from 196.72 min to 5.11 min on 128 nodes of the ORNL Frontier supercomputer. The tradeoff between accuracy and total wall-clock time is investigated and recommendations for time-parallel deployment for AM problems are made.

Additive manufacturing↗

Exploring DataVortex Systems for Irregular Applications

Emerging applications for data analytics and knowledge discovery typically have irregular or unpredictable communication patterns that do not scale well on parallel systems designed for traditional bulk-synchronous HPC applications. New network architectures that focus on minimizing (short) message latencies, rather than maximizing (large) transfer bandwidths, are emerging as possible alternatives to better support those applications with irregular communication patterns. We explore a system based upon one such novel network architecture, the Data Vortex interconnection network, and examine how this system performs by running benchmark code written for the Data Vortex network, as well as a reference MPI-over- Infiniband implementation, on the same cluster.

Data Vortex, High-Performance Computing, Networkin↗

Development of a Pulsed Slowing-Down-Time Benchmark of Neutron Thermalization in Graphite

Graphite is a classic neutronic material that has been used as both a reactor reflector and moderator in various nuclear reactor systems. The ability to accurately predict the slowing down and thermalization of neutrons in graphite can have significant implications on the safety and operation of such reactor systems. In reactors, the neutron thermalization process is quantified using the thermal scattering law (TSL) and related cross sections for a given moderator. An ideal approach to assess the validity of TSL data is using benchmark measurement based on the pulsed Slowing-Down-Time technique and its comparison with the appropriate graphite nuclear library. In this work, experimental measurement and computational Monte Carlo simulations were performed to benchmark the slowing down characteristics and thermalization of neutrons in nuclear (reactor-grade) graphite. Given the density of graphite, various graphite libraries were selected for the benchmark analysis.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Methodology for Sensitivity and Uncertainty-Based Criticality Safety Validation

Monte Carlo radiation transport software is used extensively for determining the effective multiplication factor k of fissionable systems towards assessing the criticality safety of operations. Understanding the predictive capability of the Monte Carlo software and nuclear data libraries for calculations of specific applications is vital towards making these assessments. Recently, sensitivity and uncertainty (S/U) methods have been developed for this purpose. A computational tool has been developed that uses the S/U methods with MCNP6 to select critical experiment benchmarks that have similar sources of computational bias, and extreme value theory is then used to determine suitable and conservative calculational margins. The S/U techniques are then used to perform nuclear data adjustments, which are then used to determine a suitable margin of subcriticality from cross section uncertainties. This document gives an overview of the theory of S/U methods used for criticality safety validation. First an overview is given for using the sensitivity coefficients and nuclear co variance data towards determining neutronic similarity factors is then discussed. Next, the extreme value theory (EVT) to determine a calculational margin is given along with how the similarity factors are used in the analysis. Then, the Generalized Linear Least Squares (GLLS) method for performing data adjustment and how that method is used to reject benchmarks and compute adjusted nuclear data uncertainties in k are discussed. Last, a discussion of the margin of subcriticality is given along with how to apply of the GLLS data adjustment methods to quantify the effect of nuclear data variability on k.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Simple and efficient algorithms for training machine learning potentials to force data

Machine learning models, trained on data from ab initio quantum simulations, are yielding molecular dynamics potentials with unprecedented accuracy. One limiting factor is the quantity of available training data, which can be expensive to obtain. A quantum simulation often provides all atomic forces, in addition to the total energy of the system. These forces provide much more information than the energy alone. It may appear that training a model to this large quantity of force data would introduce significant computational costs. Actually, training to all available force data should only be a few times more expensive than training to energies alone. Here, we present a new algorithm for efficient force training, and benchmark its accuracy by training to forces from real-world datasets for organic chemistry and bulk aluminum.

74 ATOMIC AND MOLECULAR PHYSICS↗

Towards interpretable Cryo-EM: disentangling latent spaces of molecular conformations

Molecules are essential building blocks of life and their different conformations (i.e., shapes) crucially determine the functional role that they play in living organisms. Cryogenic Electron Microscopy (cryo-EM) allows for acquisition of large image datasets of individual molecules. Recent advances in computational cryo-EM have made it possible to learn latent variable models of conformation landscapes. However, interpreting these latent spaces remains a challenge as their individual dimensions are often arbitrary. The key message of our work is that this interpretation challenge can be viewed as an Independent Component Analysis (ICA) problem where we seek models that have the property of identifiability. That means, they have an essentially unique solution, representing a conformational latent space that separates the different degrees of freedom a molecule is equipped with in nature. Thus, we aim to advance the computational field of cryo-EM beyond visualizations as we connect it with the theoretical framework of (nonlinear) ICA and discuss the need for identifiable models, improved metrics, and benchmarks. Moving forward, we propose future directions for enhancing the disentanglement of latent spaces in cryo-EM, refining evaluation metrics and exploring techniques that leverage physics-based decoders of biomolecular systems. Moreover, we discuss how future technological developments in time-resolved single particle imaging may enable the application of nonlinear ICA models that can discover the true conformation changes of molecules in nature. The pursuit of interpretable conformational latent spaces will empower researchers to unravel complex biological processes and facilitate targeted interventions. This has significant implications for drug discovery and structural biology more broadly. More generally, latent variable models are deployed widely across many scientific disciplines. Thus, the argument we present in this work has much broader applications in AI for science if we want to move from impressive nonlinear neural network models to mathematically grounded methods that can help us learn something new about nature.

59 BASIC BIOLOGICAL SCIENCES↗

An MLCommons Scientific Benchmarks Ontology

Scientific machine learning research spans diverse domains and data modalities, yet existing benchmark efforts remain siloed and lack standardization. This makes novel and transformative applications of machine learning to critical scientific use-cases more fragmented and less clear in pathways to impact. This paper introduces an ontology for scientific benchmarking developed through a unified, community-driven effort that extends the MLCommons ecosystem to cover physics, chemistry, materials science, biology, climate science, and more. Building on prior initiatives such as XAI-BENCH, FastML Science Benchmarks, PDEBench, and the SciMLBench framework, our effort consolidates a large set of disparate benchmarks and frameworks into a single taxonomy of scientific, application, and system-level benchmarks. New benchmarks can be added through an open submission workflow coordinated by the MLCommons Science Working Group and evaluated against a six-category rating rubric that promotes and identifies high-quality benchmarks, enabling stakeholders to select benchmarks that meet their specific needs. The architecture is extensible, supporting future scientific and AI/ML motifs, and we discuss methods for identifying emerging computing patterns for unique scientific workloads. The MLCommons Science Benchmarks Ontology provides a standardized, scalable foundation for reproducible, cross-domain benchmarking in scientific machine learning. A companion webpage for this work has also been developed as the effort evolves: https://mlcommons-science.github.io/benchmark/

Hawks, Ben [Fermilab] (ORCID:0000000157000288)↗

GASNet-EX Performance Improvements Due to Specialization for the Cray Aries Network

This document is a deliverable for milestone STPM17-6 of the Exascale Computing Project, delivered by WBS 2.3.1.14. It reports on the improvements in performance observed on Cray XC-series systems due to enhancements made to the GASNet-EX software. These enhancements, known as “specializations”, primarily consist of replacing network-independent implementations of several recently added features with implementations tailored to the Cray Aries network. Performance gains from specialization include (1) Negotiated-Payload Active Messages improve bandwidth of a ping-pong test by up to 14%, (2) Immediate Operations reduce running time of a synthetic benchmark by up to 93%, (3) non-bulk RMA Put bandwidth is increased by up to 32%, (4) Remote Atomic performance is 70% faster than the reference on a point-to-point test and allows a hot-spot test to scale robustly, and (5) non-contiguous RMA interfaces see up to 8.6x speedups for an intra-node benchmark and 26% for inter-node. These improvements are available in the GASNet-EX 2018.3.0 release.

97 MATHEMATICS AND COMPUTING↗

Simulating molecular polaritons in the collective regime using few-molecule models

The study of molecular polaritons beyond simple quantum emitter ensemble models (e.g., Tavis–Cummings) is challenging due to the large dimensionality of these systems and the complex interplay of molecular electronic and nuclear degrees of freedom. This complexity constrains existing models to either coarse-grain the rich physics and chemistry of the molecular degrees of freedom or artificially limit the description to a small number of molecules. In this work, we exploit permutational symmetries to drastically reduce the computational cost of ab initio quantum dynamics simulations for large N . Furthermore, we discover an emergent hierarchy of timescales present in these systems, that justifies the use of an effective single molecule to approximately capture the dynamics of the entire ensemble, an approximation that becomes exact as N → ∞. We also systematically derive finite N corrections to the dynamics and show that addition of k extra effective molecules is enough to account for phenomena whose rates scale as 𝒪( N − k ). Based on this result, we discuss how to seamlessly modify existing single-molecule strong coupling models to describe the dynamics of the corresponding ensemble. We call this approach collective dynamics using truncated equations (CUT-E), benchmark it against well-known results of polariton relaxation rates, and apply it to describe a universal cavity-assisted energy funneling mechanism between different molecular species. Beyond being a computationally efficient tool, this formalism provides an intuitive picture for understanding the role of bright and dark states in chemical reactivity, necessary to generate robust strategies for polariton chemistry.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A Particle-in-Cell Method for Plasmas with a Generalized Momentum Formulation, Part I: Model Formulation

Here, this paper formulates a new particle-in-cell method for the Vlasov–Maxwell system. Under the Lorenz gauge condition, Maxwell’s equations for the electromagnetic fields can be written as a collection of scalar and vector wave equations. The use of potentials for the fields motivates the adoption of a Hamiltonian formulation for particles that employs the generalized (conjugate) momentum. A notable advantage offered by the Hamiltonian formulation is the elimination of time derivatives in the Lorenz gauge formulation that are required by the standard Newton–Lorentz treatment of the particles. This allows the fields to retain the full time-accuracy guaranteed by the field solver. The resulting updates for particles require only knowledge of the fields and their spatial derivatives. An analytical method for constructing these spatial derivatives is presented that exploits the underlying integral solution used in the field solver for the wave equations. Moreover, these derivatives are demonstrated to converge at the same rate as the fields in both time and space. The Method of Lines Transpose field solver we consider in this work is globally first-order accurate in time and high-order accurate in space (e.g., fourth- and fifth-order) and belongs to a larger class of methods which are unconditionally stable, can address geometry, and leverage $\mathcal {O}(N)$ fast summation methods for efficiency. We demonstrate the method on several well-established benchmark problems on bounded domains, including a plasma sheath as well as a relativistic particle beam. The efficacy of the proposed formulation is established by comparing with a second-order accurate finite-difference time-domain method that employs a leapfrog time advance for particles and a charge conserving map suitable for bounded domains. The new method shows mesh-independent numerical heating properties even in cases where the plasma Debye length is smaller than the grid spacing. This is an important feature of the new method for problems defined on bounded domains, because it permits the use of coarser grids in space in the representation of the fields. Such a capability has significant implications for the simulation of plasmas in bounded domains with complex geometry, where the ratio between the largest and smallest cells can vary significantly. The use of high-order spatial approximations in the new method also means that fewer grid points are required in order to achieve a fixed accuracy. Our results also suggest that the new method can be used with fewer simulation particles per cell compared to the benchmark explicit method, which permits further computational savings.

97 MATHEMATICS AND COMPUTING↗

Porting AMG2013 to Heterogeneous CPU+GPU Nodes

LLNL's future advanced technology system SIERRA will feature heterogeneous compute nodes that consist of IBM PowerV9 CPUs and NVIDIA Volta GPUs. Conceptually, the motivation for such an architecture is quite straightforward: While GPUs are optimized for throughput on massively parallel workloads, CPUs strive to minimize latency for rather sequential operations. Yet, making optimal use of heterogeneous architectures raises new challenges for the development of scalable parallel software, e.g., with respect to work distribution. Porting LLNL's parallel numerical libraries to upcoming heterogeneous CPU+GPU architectures is therefore a critical factor for ensuring LLNL's future success in ful lling its national mission. One of these libraries, called HYPRE, provides parallel solvers and precondi- tioners for large, sparse linear systems of equations. In the context of this intern- ship project, I consider AMG2013 which is a proxy application for major parts of HYPRE that implements a benchmark for setting up and solving di erent systems of linear equations. In the following, I describe in detail how I ported multiple parts of AMG2013 to the GPU (Section 2) and present results for di erent experiments that demonstrate a successful parallel implementation on the heterogeneous ma- chines surface and ray (Section 3). In Section 4, I give guidelines on how my code should be used. Finally, I conclude and give an outlook for future work (Section 5).

97 MATHEMATICS AND COMPUTING↗

Bidding Curve Design for Hybrid Power Plants with Uncertain Solar Forecast

This paper presents a novel bidding curve design algorithm tailored for hybrid power plants (HPPs) to participate in the wholesale electricity market. Utilizing forecasts for photovoltaic (PV) generation and available battery power, our algorithm strategically computes the bidding curve to maximize HPP profit while adeptly managing the inherent uncertainty associated with PV power generation. In addition, the introduction of the penalty cost in HPP bidding curves provides the system operator a tool to effectively manage the system-level uncertainty that caused by HPPs. Numerical analysis through Monte Carlo simulations confirms that our bidding curve methodology outperforms the benchmark across various scenarios.

bidding curve↗

Iteration-based Linearized Distribution-level Locational Marginal Price for Three-phase Unbalanced Distribution Systems

Distributed energy resources (DERs) are rocking the utilities’ business landscape. It calls for competitive market environments that incentivize DERs to form maximum operating efficiency. Among proposed pricing schemes, distribution-level locational marginal price (DLMP) is effective in signaling the marginal generation cost differences driven by energy losses and network constraints. It can be derived from a distribution-level optimal power flow (OPF) framework, as it essentially presents the sensitivity of optimized generation cost towards incremental loads. However, due to the high resistance-to-inductance ratio and unbalanced characteristics of distribution networks, computational affordable DLMPs are highly challenged. This article provides a linear-approximated DLMP that can be solved efficiently and generalized to account for reactive power flow, three-phase unbalanced loads and meshed network structure. The successive linear programming technique is introduced to enhance the model accuracy. Case studies on an IEEE 123-Bus system validate its accuracy against a nonlinear benchmark and capability in offering proper incentives.

24 POWER TRANSMISSION AND DISTRIBUTION↗

The Hyperion Project: Partnership for an Advaned Technology Cluster Testbed

The Hyperion project offers a unique opportunity to participate in a community-driven testing and development resource at a scale beyond what can be accomplished by one entity alone. Hyperion is a new strategic technology partnership intended to support the member-driven development and testing at scale. This partnership will allow commodity clusters to scale up to meet the growing demands of customers multi-core petascale simulation environments. Hyperion will tightly couple together the outstanding research and development capabilities of Lawrence Livermore National Laboratory with leading technology companies, including Cisco, Data Direct Networks, Dell, Intel, LSI, Mellanox, Qlogic, RedHat, SuperMicro and Sun. The end goal of this project is to revolutionize cluster computing in fundamental ways by providing the critical software and hardware components for a highly scalable simulation environment. This environment will include support for high performance networking, parallel file systems, operating system, and cluster management. This goal will be achieved by building a scalable technology cluster testbed that will be fully dedicated to the partners and provide: (1) A scalable development testing and benchmarking environment for critical enabling Linux cluster technologies; (2) An evaluation testbed for new hardware and software technologies; and (3) A vehicle for forming long term collaborations.

99 GENERAL AND MISCELLANEOUS↗

Bidding Curve Design for Hybrid Power Plants with Uncertain Solar Forecast: Preprint

This paper presents a novel bidding curve design algorithm tailored for hybrid power plants (HPPs) to participate in the wholesale electricity market. Utilizing forecasts for photovoltaic (PV) generation and available battery power, our algorithm strategically computes the bidding curve to maximize HPP profit while adeptly managing the inherent uncertainty associated with PV power generation. In addition, the introduction of the penalty cost in HPP bidding curves provides the system operator a tool to effectively manage the system-level uncertainty that caused by HPPs. Numerical analysis through Monte Carlo simulations confirms that our bidding curve methodology outperforms the benchmark across various scenarios.

bidding curve↗

Manifold Learning-Based Polynomial Chaos Expansions for High-Dimensional Surrogate Models

In this work we introduce a manifold learning-based method for uncertainty quantification (UQ) in systems describing complex spatiotemporal processes. Our first objective is to identify the embedding of a set of high-dimensional data representing quantities of interest of the computational or analytical model. For this purpose, we employ Grassmannian diffusion maps, a two-step nonlinear dimension reduction technique which allows us to reduce the dimensionality of the data and identify meaningful geometric descriptions in a parsimonious and inexpensive manner. Polynomial chaos expansion is then used to construct a mapping between the stochastic input parameters and the diffusion coordinates of the reduced space. An adaptive clustering technique is proposed to identify an optimal number of clusters of points in the latent space. The similarity of points allows us to construct a number of geometric harmonic emulators which are finally utilized as a set of inexpensive pretrained models to perform an inverse map of realizations of latent features to the ambient space and thus perform accurate out-of-sample predictions. Thus, the proposed method acts as an encoder-decoder system which is able to automatically handle very high-dimensional data while simultaneously operating successfully in the small-data regime. The method is demonstrated on two benchmark problems and on a system of advection-diffusion-reaction equations which model a first-order chemical reaction between two species. In all test cases, the proposed method is able to achieve highly accurate approximations which ultimately lead to the significant acceleration of UQ tasks.

42 ENGINEERING↗

About the Impact of the Unresolved Resonance Region in Monte Carlo Simulations of Sodium Fast Reactors

In the last few years, and within the framework of different European projects, KENO-VI code from SCALE system has been employed to perform detailed continuous-energy Mon-te Carlo transport calculations for advanced fast reactors. The core characterization of both the sodium-cooled ASTRID and the lead-cooled ALFRED reactors was performed during the FP7 cross-cutting ESNII+ project; more recently, core calculations for the sodium-cooled Superphénix reactor and the improved European Sodium Fast Reactor design were performed within the HORIZON2020 ESFR-SMART project. In all cases, the effective multiplication factor predicted by KENO-VI was systematically higher (around 400-500 pcm) than the values computed by MCNP and Serpent Monte Carlo codes, using the same nu-clear data library.In order to provide insight into the origin of the observed discrepancies, a simplified 2D MOX-fueled SFR pin-cell benchmark has been launched. The multiplication factor, as well as 1-group and VITAMINJ 175-group cross-sections computed by KENO-VI, Serpent and MCNP codes employing ENDF/B-VII.1 data library, have been compared.Significant differences between KENO-VI and the other codes have been found in the un-resolved resonance regions of 239Pu and 241Pu capture and production cross sections, while negligible differences appeared outside those energy ranges. On the other hand, calculations without using probability tables have shown very good agreement. Quantita-tive comparison is presented and analyzed, along with a discussion of the impact of the probability-table treatment in the three codes for MOX-fueled systems with typical SFR spectrum.

Jimenez carrascosa, A↗