Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallel computers.”

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 1,513 records · Page 84

Estimate the Electromechanical States Using Particle Filtering and Smoothing

Accurate knowledge of electromechanical states is critical for efficient and reliable control of a power system. This paper proposes a particle filtering approach to estimate the electromechanical states of power systems from Phasor Measurement Unit (PMU) data. Without having to go through laborious linearization procedure, the proposed particle filtering techniques can estimate states of a complex power system, which is often non-linear and has non-Gaussian noise. The proposed method is evaluated using a multi-machine system with both large and small disturbances. Sensitivity studies of the dynamic state estimation performance are also presented to show the robustness of the proposed method. The inherent decoupling properties of particle filtering make it highly scalable and the potential to reduce computational time through parallel implementation is very promising.

Auxiliary Particle Filter↗

Development of a Stiffness-Based Chemistry Load Balancing Scheme, and Optimization of Input/Output and Communication, to Enable Massively Parallel High-Fidelity Internal Combustion Engine Simulations

A closed-cycle gasoline compression ignition engine simulation near top dead center (TDC) was used to profile the performance of a parallel commercial engine computational fluid dynamics code, as it was scaled on up to 4096 cores of an IBM Blue Gene/Q supercomputer. The test case has 9 million cells near TDC, with a fixed mesh size of 0.15 mm, and was run on configurations ranging from 128 to 4096 cores. Profiling was done for a small duration of 0.11 crank angle degrees near TDC during ignition. Optimization of input/output performance resulted in a significant speedup in reading restart files, and in an over 100-times speedup in writing restart files and files for post-processing. Improvements to communication resulted in a 1400-times speedup in the mesh load balancing operation during initialization, on 4096 cores. An improved, “stiffness-based” algorithm for load balancing chemical kinetics calculations was developed, which results in an over 3-times faster run-time near ignition on 4096 cores relative to the original load balancing scheme. With this improvement to load balancing, the code achieves over 78% scaling efficiency on 2048 cores, and over 65% scaling efficiency on 4096 cores, relative to 256 cores.

CFD, Mira, Supercomputing, High Performance Comput↗

Influence of Vane Trailing Edge Flow on the Formation of Cavity Cells and Rim Sealing

Abstract Hot gas ingestion into the turbine rim seal cavity is an important concern for engine designers. To prevent ingestion, rim seals use high-pressure purge flow; however, the penalty is that excessive use of the purge flow decreases engine thermal efficiency. In this paper, a one-stage turbine operating at engine-representative conditions was used to study the effect of steady and time-resolved under-platform cavity temperatures and pressures across a range of coolant flowrates in the presence of vane trailing edge (VTE) flow. This study correlates time-resolved pressure with time-resolved temperature to identify primary frequencies driving ingestion. At certain flowrates, the time-resolved pressures are out of phase with the temperatures, indicating ingestion. Measurements from high-frequency response pressure sensors in the rim seal and vane platform were also used to determine rotational speed and quantity of large-scale structures (cells). In a parallel effort, a computational model using Unsteady Reynolds-averaged Navier–Stokes (URANS) was applied to determine swirl ratio in the rim seal cavity and time-resolved rim sealing effectiveness. The experimental results confirm that at low purge flowrates, the VTE flow influences the unsteady flow field by decreasing pressure unsteadiness in the rim seal cavity. Results show an increase in purge flow increases the number of unsteady large-scale structures in the rim seal and decreases their rotational speed. However, VTE flow was shown to not significantly change the cell speed and count in the rim seal. Simulations point to the importance of the large-scale cell structures in influencing rim sealing unsteadiness, which is not captured in current rim sealing predictive models.

Engineering↗

A high-performance parallel algorithm for nonnegative matrix factorization

Non-negative matrix factorization (NMF) is the problem of determining two non-negative low rank factors W and H, for the given input matrix A, such that A ≈ WH. NMF is a useful tool for many applications in different domains such as topic modeling in text mining, background separation in video analysis, and community detection in social networks. Despite its popularity in the data mining community, there is a lack of efficient distributed algorithms to solve the problem for big data sets. We propose a high-performance distributed-memory parallel algorithm that computes the factorization by iteratively solving alternating non-negative least squares (NLS) subproblems for W and H. It maintains the data and factor matrices in memory (distributed across processors), uses MPI for interprocessor communication, and, in the dense case, provably minimizes communication costs (under mild assumptions). As opposed to previous implementations, our algorithm is also flexible: (1) it performs well for both dense and sparse matrices, and (2) it allows the user to choose any one of the multiple algorithms for solving the updates to low rank factors W and H within the alternating iterations. We demonstrate the scalability of our algorithm and compare it with baseline implementations, showing significant performance improvements.

Kannan, Ramakrishnan↗

AGS-GNN: Attribute-guided Sampling for Graph Neural Networks

We propose AGS-GNN, a novel attribute-guided sampling algorithm for Graph Neural Networks (GNNs) that exploits node features and connectivity structure of a graph while simultaneously adapting for both homophily and heterophily in graphs. (In homophilic graphs vertices of the same class are more likely to be connected, and vertices of different classes tend to be linked in heterophilic graphs.) While GNNs have been successfully applied to homophilic graphs, their application to heterophilic graphs remains challenging. The best-performing GNNs for heterophilic graphs do not fit the sampling paradigm, suffer high computational costs, and are not inductive. We employ samplers based on feature-similarity and feature-diversity to select subsets of neighbors for a node, and adaptively capture information from homophilic and heterophilic neighborhoods using dual channels. Currently, AGS-GNN is the only algorithm that we know of that explicitly controls homophily in the sampled subgraph through similar and diverse neighborhood samples. For diverse neighborhood sampling, we employ submodularity, which was not used in this context prior to our work. The sampling distribution is pre-computed and highly parallel, achieving the desired scalability. Using an extensive dataset consisting of 35 small (<=100K nodes) and large (>100K nodes) homophilic and heterophilic graphs, we demonstrate the superiority of AGS-GNN compare to the current approaches in the literature. AGS-GNN achieves comparable test accuracy to the best-performing heterophilic GNNs, even outperforming methods using the entire graph for node classification. AGS-GNN also converges faster compared to methods that sample neighborhoods randomly, and can be incorporated into existing GNN models that employ node or graph sampling.

artificial intelligence↗

Performance Models for the Spike Banded Linear System Solver

With availability of large-scale parallel platforms comprised of tens-of-thousands of processors and beyond, there is significant impetus for the development of scalable parallel sparse linear system solvers and preconditioners. An integral part of this design process is the development of performance models capable of predicting performance and providing accurate cost models for the solvers and preconditioners. There has been some work in the past on characterizing performance of the iterative solvers themselves. In this paper, we investigate the problem of characterizing performance and scalability of banded preconditioners. Recent work has demonstrated the superior convergence properties and robustness of banded preconditioners, compared to state-of-the-art ILU family of preconditioners as well as algebraic multigrid preconditioners. Furthermore, when used in conjunction with efficient banded solvers, banded preconditioners are capable of significantly faster time-to-solution. Our banded solver, the Truncated Spike algorithm is specifically designed for parallel performance and tolerance to deep memory hierarchies. Its regular structure is also highly amenable to accurate performance characterization. Using these characteristics, we derive the following results in this paper: (i) we develop parallel formulations of the Truncated Spike solver, (ii) we develop a highly accurate pseudo-analytical parallel performance model for our solver, (iii) we show excellent predication capabilities of our model – based on which we argue the high scalability of our solver. Our pseudo-analytical performance model is based on analytical performance characterization of each phase of our solver. These analytical models are then parameterized using actual runtime information on target platforms. An important consequence of our performance models is that they reveal underlying performance bottlenecks in both serial and parallel formulations. All of our results are validated on diverse heterogeneous multiclusters – platforms for which performance prediction is particularly challenging. Finally, we provide predict the scalability of the Spike algorithm using up to 65,536 cores with our model. In this paper we extend the results presented in the Ninth International Symposium on Parallel and Distributed Computing.

Manguoglu, Murat↗

JBlulce Data Acquisition Software for Macromolecular Crystallography

JBlulce (Java Beam Line Universal Integrated Configuration Environment is a data acquisition software for macromolecular crystallography conforming user interface of the SSRL Blulce that has become a de-factor standard in the field. Besides this interface conformity, JBlulce is a unique system in terms of architecture, spec, capability and software implementation. It features only two software layers, the JBlulce clients and the EPICS servers, as compared to three layers present in Blulce and most of similar systems. This layers reduction provides a faster communication with hardware and an easier access to advanced hardware capabilities like on-the-fly scanning. Then JBlulce clients are designed to operate in parallel with the other beamline controls which streamlines the tasks performed by staff such as beamline preparation, maintenance, auditing and user assistance. Another distinction is the deployment of multiple plugins that can be written in any programming language thus involving more staff into the development. further on, JBlulce makes use of unified motion controls allowing for easy scanning and optimizing of any beamline component. Finally, the graphic interface is implemented in Java making full use of rich Java libraries and Jave IDE for debugging. to compare, Blulce user interface is implemented with aging Tcl/tk language providing very restricted capabilities. JBlulce makes full use of the industrial power and wide drivers selection of EPICS in controlling hardware; all hardware communication is routed via multiple EPICS servers residing on local area network. JBlulce also includes several EPICS State Notation servers aimed at making hardware communication more robust. Besides using EPICS for controlling hardware, JBlulce extensively uses EPICS databases for efficient communications between multiple instances of JBlulce clients and JBlulce plugins that can run in parallel on different computers. All of the above makes JBlulce one of the biggest and most sophisticated EPICS client projects to date. JBlulce configuration is stored in my SQL database which provides flexibility in tuning the system. The database is also accessible by the plugins. From the users perspective JBlulce provides all standard features of data acquisition software for macromolecular crystallography plus such unique capabilities as one click beamline energy change that may involve switching undulator harmonics, mirrors lanes and beam realignment, automated diffraction rtastering for finding small crystals and sweet spots on poorly diffracting crystals with automated scoring of raster cells by the number of reflections; data collection along a vector; automated on-the-fly fluorescent tastering, a faster and lower-irradiation compliment to the diffraction raster; fully automated fluorescence measurements for MAD that include signal optimization, fast on the fly energy scanning and automated adapting of scan range to chemical shifts; fly-scan minibeam realignment; automated loop and crystal centering, controls for sample automounter, automated screening, data collection auditing, remote access and a lot more.

Stepanov, Sergey↗

HPC-NMF: A High-Performance Parallel Algorithm for Nonnegative Matrix Factorization

NMF is a useful tool for many applications in different domains such as topic modeling in text mining, background separation in video analysis, and community detection in social networks. Despite its popularity in the data mining community, there is a lack of efficient distributed algorithms to solve the problem for big data sets. We propose a high-performance distributed-memory parallel algorithm that computes the factorization by iteratively solving alternating non-negative least squares (NLS) subproblems for $\WW$ and $\HH$. It maintains the data and factor matrices in memory (distributed across processors), uses MPI for interprocessor communication, and, in the dense case, provably minimizes communication costs (under mild assumptions). As opposed to previous implementation, our algorithm is also flexible: It performs well for both dense and sparse matrices, and allows the user to choose any one of the multiple algorithms for solving the updates to low rank factors $\WW$ and $\HH$ within the alternating iterations.

Kannan, Ramakrishnan↗

MARBLES (Multi-scale Adaptively Refined Boltzmann LatticE Solver) [SWR-23-37]

MARBLES (Multi-scale Adaptively Refined Boltzmann LatticE Solver) is an open-source computational fluid dynamics package powered by the lattice Boltzmann equations and built on AMReX. In the lattice Boltzmann method, local collisions between meso-scale fictitious particles drive the governing equations which enables MARBLES to easily simulate flow around complex and/or moving geometry without the generation of a body-conforming mesh. Using AMReX data structures and operations ensures a high level of computational performance and parallel scaling on heterogenous architectures while also naturally supporting locally enhanced grid resolution and fidelity through automatic mesh refinement. New domains and problem definitions are easily specified through an input file with examples and guidance on all options and variables provided in the MARBLES documentation.

Henry de Frahan, Marc↗

PFLOTRAN 5

PFLOTRAN leverages massively parallel, high performance computing to simulate large-scale non-isothermal multiphase flow, multicomponent reactive transport and electrical resistivity tomography (ERT) problems in the subsurface environment. Researchers have employed PFLOTRAN to simulate these Earth system processes on leadership class supercomputers for over two decades. The code is designed to predict the future estate of environmental systems and better inform stakeholders in the regulatory decision making process (e.g., fate of contaminants, long-term stewardship for nuclear waste, impact of climate change, etc.). A diverse team of scientists oversees PFLOTRAN development and maintenance under an open-source licensing agreement and manages contributions from an international community of researchers.

Hammond, Glenn↗

LLM Benchmarking with LLaMA2: Evaluating Code Development Performance Across Multiple Programming Languages

The rapid evolution of large language models (LLMs) has opened new possibilities for automating various tasks in software development. This paper evaluates the capabilities of the LLaMA 2-70B model in automating these tasks for scientific applications written in commonly used programming languages. Using representative test problems, we assess the model's capacity to generate code, documentation, and unit tests, as well as its ability to translate existing code between commonly used programming languages. Our comprehensive analysis evaluates the compilation, runtime behavior, and correctness of the generated and translated code. Additionally, we assess the quality of automatically generated code, documentation, and unit tests. Here, our results indicate that while LLaMA 2-70B frequently generates syntactically correct and functional code for simpler numerical tasks, it encounters substantial difficulties with more complex, parallelized, or distributed computations, requiring considerable manual corrections. We identify key limitations and suggest areas for future improvements to better leverage AI-driven automation in scientific computing workflows.

97 MATHEMATICS AND COMPUTING↗

Highlights of X-Stack ExM Deliverable: MosaStore

This brief report highlights the experience gained with MosaStore, an exploratory part of the X-Stack project “ExM: System support for extreme-scale, many-task applications”. The ExM project proposed to use concurrent workflows supported by the Swift language and runtime as an innovative programming model to exploit parallelism in exascale computers. MosaStore aims to support this endeavor by improving storage support for workflow-based applications, more precisely by exploring the gains that can be obtained from co-designing the storage system and the workflow runtime engine. MosaStore has been developed primarily at the University of British Columbia.

97 MATHEMATICS AND COMPUTING↗

Entity Resolution at Large Scale: Benchmarking and Algorithmics

We seek scalable benchmarks for entity resolution problems. Solutions to these problems range from trivial approaches such as string sorting to sophisticated methods such as statistical relational learning. The theoretical and practical complexity of these approaches varies widely, so one of the primary purposes of a benchmark will be to quantify the trade-off between solution quality and runtime. We are motivated by the ubiquitous nature of entity resolution as a fundamental problem faced by any organization that ingests large amounts of noisy text data. A benchmark is typically a rigid specification that provides an objective measure usable for ranking implementations of an algorithm. For example the Top500 and HPCG500 bench- marks rank supercomputers based on their performance of dense and sparse linear algebra problems (respectively). These two benchmarks require participants to report FLOPS counts attainable on various machines. Our purpose is slightly different. Whereas the supercomputing benchmarks mentioned above hold algorithms constant and aim to rank machines, we are primarily interested in ranking algorithms. As mentioned above, entity resolution problems can be approached in completely different ways. We believe that users of our benchmarks must decide what sort of procedure to run before comparing implementations and architectures. Eventually, we also wish to provide a mechanism for ranking machines while holding algorithmic approach constant . Our primary contributions are parallel algorithms for computing solution quality mea- sures per entity. We find in some real datasets that many entities are quite easy to resolve while others are difficult, with a heavy skew toward the former case. Therefore, measures such as global confusion matrices, F measures, etc. do not meet our benchmarking needs. We design methods for computing solution quality at the granularity of a single entity in order to know when proposed solutions do well in difficult situations (perhaps justifying extra computational), or struggling in easy situations. We report on progress toward a viable benchmark for comparing entity resolution algo- rithms. Our work is incomplete, but we have designed and prototyped several algorithms to help evalute the solution quality of competing approaches to these problems. We envision a benchmark in which the objective measure is a ratio of solution quality to runtime.

97 MATHEMATICS AND COMPUTING↗

HIPPO – A Software Platform for Electricity Market Research and Development

The goal of this project is to provide Regional transmission organizations (RTOs) and independent system operators (ISOs) a market design and prototyping software, High-Performance Power-Grid Optimization (HIPPO), that they can evaluate electricity market design options, calculate market planning strategies and operational performance. With the high standards and strict reliability requirements for operating power systems, impacts of new technologies need to be fully investigated prior to any consideration for adoption. A market design and prototyping software tool which can be used to prototype electricity market design options, to calculate market planning strategies and operational performance with high precision, and to investigate the impacts for integrating future power grid technologies will be valuable to RTOs/ISOs who operate power systems, to vendors like GE and ABB who provide the market solvers, and to market participants and researchers who are actively doing market research. HIPPO is a such tool that can be used to improve the current market operations and provide capabilities for rigorous forward-looking design and prototyping of next-generation energy markets. HIPPO has a high-resolution model for the day-ahead SCUC, which was validated with MISO and GE-Grid Solutions. HIPPO is built with parallel and distributed computing capabilities and can be executed in both multi-thread and high-performance computing (HPC) settings. This capability provides fast solution speed necessary to handle the larger and more complex SCUC problems of real-world cases and the potentially growing size and complexity of future scenarios. In addition, HIPPO has a concurrent optimizer (CO) which manages multiple algorithm executions simultaneously and leverages the advantages from different algorithms. This structure provides flexibility to better benchmark competing approaches. Highly accurate market model, fast solution technologies and flexible model and algorithm control are the features which will make HIPPO an extensible platform for developing and testing multiple approaches to meet a wide range of future market needs.

24 POWER TRANSMISSION AND DISTRIBUTION↗

MultiPEM Toolbox: User Manual [Rev. 2]

This document explains use of the Multi-Phenomenology Explosion Monitoring (Multi PEM) Toolbox, a collection of R scripts for estimating the unknown device parameters of a new event with uncertainty quantification. The methodology and application used for illustration in this user manual are fully documented in a Los Alamos National Laboratory technical report hereafter designated “WPA” for reference. Additional details on the application are found in a recent journal article. Two assessment types are available: rapid and complete. Rapid assessments are conducted in two stages, as described in Section 2. In the first stage, calibration data are used to estimate forward and error model parameters (WPA, §5.1) and (if relevant) errors-in-variables yield values for calibration sources (WPA, §3, Equation (3)). In the second stage, new event data are used to estimate the unknown new event device parameters (WPA, §5.2) with uncertainty quantification. Two options for treating the inferred first stage parameters in second stage Bayesian analysis are available: fixing them at their maximum likelihood estimate (default), or multiple imputation. Multiple imputation involves utilizing several posterior samples (imputations) of the first stage parameters as fixed values in the second stage posterior sampling of the new event device parameters. Second stage sampling is conducted across imputations in parallel to improve computational efficiency. This method produces improved uncertainty quantification of the new event device parameters compared with the default treatment of the first stage parameters, at the expense of additional computation. Complete assessments are conducted in a single stage, as described in Section 3. Calibration and (if relevant) new event data are used simultaneously to estimate all forward model, error model, and (if relevant) new event device parameters with uncertainty quantification on the latter. As the name suggests, rapid assessments generally run substantially faster than complete assessments (even with multiple imputation), because the results of first stage analysis can be stored and incorporated into estimating a relatively low-dimensional space of new event device parameters whenever relevant new event data becomes available. On the other hand, complete assessments must be run on the full set of model and device parameters with calibration and new event data every time the latter becomes available.

97 MATHEMATICS AND COMPUTING↗

FY08 LDRD Final Report A New Method for Wave Propagation in Elastic Media LDRD Project Tracking Code: 05-ERD-079

The LDRD project 'A New Method for Wave Propagation in Elastic Media' developed several improvements to the traditional finite difference technique for seismic wave propagation, including a summation-by-parts discretization which is provably stable for arbitrary heterogeneous materials, an accurate treatment of non-planar topography, local mesh refinement, and stable outflow boundary conditions. This project also implemented these techniques in a parallel open source computer code called WPP, and participated in several seismic modeling efforts to simulate ground motion due to earthquakes in Northern California. This research has been documented in six individual publications which are summarized in this report. Of these publications, four are published refereed journal articles, one is an accepted refereed journal article which has not yet been published, and one is a non-refereed software manual. The report concludes with a discussion of future research directions and exit plan.

58 GEOSCIENCES↗

Limits on the Efficiency of Event-Based Algorithms for Monte Carlo Neutron Transport

The traditional form of parallelism in Monte Carlo particle transport simulations, wherein each individual particle history is considered a unit of work, does not lend itself well to data-level parallelism. Event-based algorithms, which were originally used for simulations on vector processors, may offer a path toward better utilizing data-level parallelism in modern computer architectures. In this study, a simple model is developed for estimating the efficiency of the event-based particle transport algorithm under two sets of assumptions. Data collected from simulations of four reactor problems using OpenMC was then used in conjunction with the models to calculate the speedup due to vectorization as a function of two parameters: the size of the particle bank and the vector width. When each event type is assumed to have constant execution time, the achievable speedup is directly related to the particle bank size. We observed that the bank size generally needs to be at least 20 times greater than vector size in order to achieve vector efficiency greater than 90%. When the execution times for events are allowed to vary, however, the vector speedup is also limited by differences in execution time for events being carried out in a single event-iteration. For some problems, this implies that vector effciencies over 50% may not be attainable. While there are many factors impacting performance of an event-based algorithm that are not captured by our model, it nevertheless provides insights into factors that may be limiting in a real implementation.

Romano, Paul K.↗

A Methodology for the Rapid Development of Scalable HPC Data Services

There is growing evidence in the scientifc computing community that parallel file systems are not sufficient for all HPC storage workloads. This realization has motivated extensive research in new storage system designs. The question of which alternative we should turn towards implies that there could be a single answer satisfying a wide range of very diverse applications. We argue that such a generic solution does not exist. Instead, custom data services should be designed and tailored to the needs of specific applications on specific hardware. Furthermore, custom data services should be designed in close collaboration with users. In this paper, we present a methodology for the rapid development of such data services. This methodology promotes the design of reusable building blocks that can be composed together efficiently through a runtime based on high-performance threading, tasking, and remote procedure calls. We illustrate the success of our methodology by showcasing three examples of data services designed from the same building blocks, yet targeting entirely different applications.

Dorier, Mattieu↗