SEARCH · Engineering Papers

Results for “parallel and distributed computing”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,045 records · Page 58

All sources NASA NTRS DOE OSTI

DOE OSTI2021

MFIX DEM Enhancement for Industry-Relevant Flows (Final Report)

The overall goal of this two-phase project is to implement performance improvements of the Multiphase Flow with Interphase Exchanges (MFIX) Discrete Element Model (DEM) code that enable a transformative shift for industrial use. Prior to this effort, the largest simulations performed using MFIX are O(10 7 ) particles. This falls short of the O(10 9 ) particle simulations that must be completed on a timescale of days or weeks (vs. months or years) to enable simulations with physically-relevant domain sizes to be incorporated into industrial design cycles within five years. This was accomplished by tailoring best-in-class practices to bear on the unique challenges posed by the MFIX-DEM algorithm and code base. Scientific simulations (e.g., in cosmology, turbulent combustion) routinely use massively parallel computing to update far more particles in short wall clock times. Results from Phase 1 (1.5 years in duration) indicated significant gains in speed were possible for a wide range of benchmark cases. Moreover, a survey sent to >35 companies indicates that the timing is ideal for such an enhanced tool, with >80% of the respondents indicating that DEM is already value-added or will be within the next 5 years, and >70% of the respondents indicating that improved speed is the top computational priority. In Phase 2 (3.5 years in duration), the two major barriers that hinder industry from effectively using multiphase Computational Fluid Dynamics (CFD) to cut costs and improve performance, namely computational overhead and confidence in predictions, continued to be addressed. Regarding the former, the results from Phase 1 to guide the effort, with enhancements focused on an improved time-stepping algorithm and particle sorting. Four target problems of 1 billion particles each and increasing complexity were identified: homogeneous cooling, tumbler with continuous particle size distribution, discharge from a rectangular hopper and a cylindrical riser. Each of these were successfully simulated for relevant time scales (on order of seconds) using less than 24 hours of wall clock time. These represent the first 1-billion particle DEM simulations performed with MFIX, namely using the MFIX-Exa code. This code is currently under development at NETL in collaboration with Lawrence Berkeley National Laboratory. Regarding the second barrier on predictive uncertainty, experiments from Phase 1 (interacting nozzles - hydrodynamics only) and Phase 2 (very small-scale segregation experiments) were used to demonstrate the ability of two simplified approaches to uncertainty quantification (UQ). By limiting the number of particles, UQ based on the simplified treatment was compared to standard UQ, which was shown to have much higher computational demands. Experiments were also performed on a pilot-scale stripper unit to provide validation data for future CFD-DEM simulations and UQ.

20 FOSSIL-FUELED POWER PLANTS↗

NASA NTRS1990

Control of multiple resonant power processors in a multi-source system

Analysis and test results show that phasor-regulated, Mapham-derived resonant inverters can be paralleled to provide standardizing interfaces for multiple sources on a utility-type, aerospace power distribution bus. The basic sources do not require matching in any way, and may have grossly different characteristics. Fully stable system architectures with multiple sources, parallel/redundant distribution buses, and a wide variety of loads can be easily constructed and controlled. The commands and parameters available for system control allow for tight tolerance bus voltage control, and absolute power-sharing control from the various sources over the full range of possible source and load variations. That level of control enables simplified load power processing hardware and the distribution of losses to optimally load the source thermal control system. Positive control of all system performance and allocation of losses are not required by all missions or vehicles, and overall vehicle considerations do not always require the loads on vehicle energy sources and thermal control systems to be balanced. In those cases, power system control can be simplified, and a hierarchical set of defaults can be substituted for computer-generated or supervisory input commands to allow for stable, fully autonomous system operation.

Mildice, James↗

DOE OSTI2024

Impact of Reordering on the LU Factorization Performance of Bordered Block-Diagonal Sparse Matrix

Power engineers rely on computer-based simulation tools to assess grid performance and ensure security. At the core of these tools are solvers for sparse linear equations. When transformed into a bordered block-diagonal (BBD) structure, part of the sparse linear equation solving can be parallelized. This work focuses on using the Schur-complement-based method for LU factorization on BBD matrices, specifically, Jacobian matrices from large-scale systems. Our findings show that the natural ordering method outperforms the default ordering method in computational performance for each block of the BBD matrix. This observation is validated using synthetic 25k-bus and 70k-bus cases, showing a speedup of up to 38% when using natural ordering without permutation. Additionally, the impact of the number of partitions is studied, and the result shows that computational performance improves with more, smaller partitions in the BBD matrices.

BBD matrix↗

DOE OSTI2024

Parallel Runtime Interface for Fortran (PRIF) Specification (Rev. 0.5)

This document specifies an interface to support the parallel features of Fortran, named the Parallel Runtime Interface for Fortran (PRIF). PRIF is a proposed solution in which the runtime library is primarily responsible for implementing coarray allocation, deallocation and accesses, image synchronization, atomic operations, events, teams and collective subroutines. In this interface, the compiler is responsible for transforming the invocation of Fortran-level parallel features into procedure calls to the necessary PRIF subroutines. The interface is designed for portability across shared- and distributed-memory machines, different operating systems, and multiple architectures. Implementations of this interface are intended as an augmentation for the compiler's own runtime library. With an implementation-agnostic interface, alternative parallel runtime libraries may be developed that support the same interface. One benefit of this approach is the ability to vary the communication substrate. A central aim of this document is to define a parallel runtime interface in standard Fortran syntax, which enables us to leverage Fortran to succinctly express various properties of the procedure interfaces, including argument attributes.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1993

High-performance mass storage system for workstations

Reduced Instruction Set Computer (RISC) workstations and Personnel Computers (PC) are very popular tools for office automation, command and control, scientific analysis, database management, and many other applications. However, when using Input/Output (I/O) intensive applications, the RISC workstations and PC's are often overburdened with the tasks of collecting, staging, storing, and distributing data. Also, by using standard high-performance peripherals and storage devices, the I/O function can still be a common bottleneck process. Therefore, the high-performance mass storage system, developed by Loral AeroSys' Independent Research and Development (IR&D) engineers, can offload a RISC workstation of I/O related functions and provide high-performance I/O functions and external interfaces. The high-performance mass storage system has the capabilities to ingest high-speed real-time data, perform signal or image processing, and stage, archive, and distribute the data. This mass storage system uses a hierarchical storage structure, thus reducing the total data storage cost, while maintaining high-I/O performance. The high-performance mass storage system is a network of low-cost parallel processors and storage devices. The nodes in the network have special I/O functions such as: SCSI controller, Ethernet controller, gateway controller, RS232 controller, IEEE488 controller, and digital/analog converter. The nodes are interconnected through high-speed direct memory access links to form a network. The topology of the network is easily reconfigurable to maximize system throughput for various applications. This high-performance mass storage system takes advantage of a 'busless' architecture for maximum expandability. The mass storage system consists of magnetic disks, a WORM optical disk jukebox, and an 8mm helical scan tape to form a hierarchical storage structure. Commonly used files are kept in the magnetic disk for fast retrieval. The optical disks are used as archive media, and the tapes are used as backup media. The storage system is managed by the IEEE mass storage reference model-based UniTree software package. UniTree software will keep track of all files in the system, will automatically migrate the lesser used files to archive media, and will stage the files when needed by the system. The user can access the files without knowledge of their physical location. The high-performance mass storage system developed by Loral AeroSys will significantly boost the system I/O performance and reduce the overall data storage cost. This storage system provides a highly flexible and cost-effective architecture for a variety of applications (e.g., realtime data acquisition with a signal and image processing requirement, long-term data archiving and distribution, and image analysis and enhancement).

Chiang, T.↗

DOE OSTI2024

3D modeling of deep borehole electromagnetic measurements with energized casing source for fracture mapping at the Utah Frontier Observatory for Research in Geothermal Energy

Here, we present a 3D numerical modelling analysis evaluating the deployment of a borehole electromagnetic measurement tool to detect and image a stimulated zone at the Utah Frontier Observatory for Research in Geothermal Energy geothermal site. As the depth to the geothermal reservoir is several kilometres and the size of the stimulated zone is limited to several 100 m, surface-based controlled-source electromagnetic measurements lack the sensitivity for detecting changes in electrical resistivity caused by the stimulation. To overcome the limitation, the study evaluates the feasibility of using a three-component borehole magnetic receiver system at the Frontier Observatory for Research in Geothermal Energy site. To provide sufficient currents inside and around the enhanced geothermal reservoir, we use an injection well as an energized casing source. To efficiently simulate energizing the injection well in a realistic 3D resistivity model, we introduce a novel modelling workflow that leverages the strengths of both 3D cylindrical-mesh-based electromagnetic modelling code and 3D tetrahedral-mesh-based electromagnetic modelling code. The former is particularly well-suited for modelling hollow cylindrical objects like casings, whereas the latter excels at representing more complex 3D geological structures. In this workflow, our initial step involves computing current densities along a vertical steel-cased well using a 3D cylindrical electromagnetic modelling code. Subsequently, we distribute a series of equivalent current sources along the well's trajectory within a complex 3D resistivity model. We then discretize this model using a tetrahedral mesh and simulate the borehole electromagnetic responses excited by the casing source using a 3D finite-element electromagnetic code. This multi-step approach enables us to simulate 3D casing source electromagnetic responses within a complex 3D resistivity model, without the need for explicit discretization of the well using an excessive number of fine cells. We discuss the applicability and limitations of this proposed workflow within an electromagnetic modelling scenario where an energized well is deviated, such as at the Frontier Observatory for Research in Geothermal Energy site. Using the workflow, we demonstrate that the combined use of the energized casing source and the borehole electromagnetic receiver system offer measurable magnetic field amplitudes and sensitivity to the deep localized stimulated zone. The measurements can also distinguish between parallel-fracture anisotropic reservoirs and isotropic cases, providing valuable insights into the fracture system of the stimulated zone. Besides the magnetic field measurements, vertical electric field measurements in the open well sections are also highly sensitive to the stimulated zone and can be used as additional data for detecting and imaging the target. We can also acquire additional multiple-source data by grounding the surface electrode at various locations and repeating borehole electromagnetic measurements. This approach can increase the number of monitoring data by several factors, providing a more comprehensive dataset for analysing the deep-localized stimulated zone. The numerical analysis indicates that it is feasible to use the combination of the energized casing and downhole electromagnetic measurements in monitoring localized stimulated zone at large depths.

58 GEOSCIENCES↗

DOE OSTI2021

How the Size and Density of Charge-Transfer Excitons Depend on Heterojunction’s Architecture

We have characterized the size, intensity, density, and distribution of charge-transfer (CT) excitons as a function of the acceptor–donor architecture of prototypical organic interfaces. This characterization was done by computational analysis of 17 models of varying numbers, positions, and orientations of the donor and acceptor molecules. The models’ building blocks were phenyl-C61-butyric acid methyl ester (PCBM) fullerene acceptors and dual-band donor polymers composed of thiophene, benzothiadiazole, and benzotriazole subunits. The electronic structure of the donor–acceptor complexes was computed with the time-dependent long-range-corrected density-functional tight-binding method and analyzed with the fragment-based one-electron transition density matrix. In all models, the complexes with edge-on orientation have denser spectra of low-energy CT states lying below the absorption bands compared to the complexes with face-on orientation. This CT-state distribution in edge-on complexes provides a gate to efficiently populate cold CT excitons. Moreover, the cold CT excitons have a higher degree of charge separation in the edge-on than in the face-on complexes. The CT amount and the CT exciton size generally increase with the energy of the CT states, although the electron remains localized on a single molecule in cold CT states. Delocalization over two PCBM molecules was observed for high-energy CT states. The exciton size also depends on the orientation. Larger excitons are produced by the delocalization of the electrons perpendicularly to the interface. When the delocalization is parallel, the smaller electron–hole distances yield moderately sized CT excitons.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2021

Progress on Optimizing Wind Farms and Rotor Designs Using Adjoints

Modern wind plants are increasingly tasked with multiple performance objectives. In addition to designing plants that maximize power output and minimize the levelized cost of energy (LCOE), the design and operation of wind plants is increasingly influenced by challenges regarding grid integration of variable generation renewables. This places a growing emphasis on making wind plants more controllable and predictable. WindSE is a Reynolds-averaged Navier-Stokes (RANS) model designed around analytical gradient and adjoint methods, with the ability to capture terrain-induced effects, as shown in Figure 1. The recent addition of an unsteady solver with an actuator line method (ALM) and ongoing work to enable massively parallel optimizations gives it a unique niche to explore coupled plant-level controls and design problems. This code is an open source python package built on the FEniCS framework that utilizes fast, parallel PETSc solvers to model fluid flow throughout wind-farm scale domains. Two recent studies performed using WindSE demonstrate the capability to optimize under a wide variety of flow conditions and objective functions. In the first, we present an optimization focused on modifying the layout of a wind farm with a fixed number of turbines for maximum total power output [1]. This study highlights the ability to quickly perform simulations using the steady Navier-Stokes solver combined with rotors represented as actuator disks while also stressing the importance of capturing terrain-induced effects. Gradient-based optimization using the RANS equations is viable due to the inclusion of efficiently computed adjoint derivatives. We interpret the physical results of the optimal layout and also discuss the computational cost of scaling to larger problems. In the second study, we present the capabilities of the unsteady Navier-Stokes solver, where rotor-blade profiles represented by actuator lines are optimized to enhance wake steering effects and overall power production [2]. We quantify the wind plant performance gains obtained from this type of simultaneous control co-design optimization as compared to optimizing the blade design and yaw independently. Figure 2 shows the differences between a baseline two-turbine system and an optimized system where we fine-tune the blade chord profile. Results and challenges from each study are quickly summarized and used to motivate the current development efforts within WindSE. Current and future work is focused on enabling higher-resolution studies with more degrees of freedom through parallelization of both the simulation and optimization algorithms. We present benchmarking results to show that WindSE performs well in both weak- and strong-scaling tests and further demonstrate that the optimizer obtains the same convergence rates in both shared- and distributed-memory environments. Using larger wind farms, we can study deep-array effects within an optimization context, allowing the use of objective functions that have been previously unstudied. As an example, we present ongoing work on a blockage metric which characterizes the loss of available kinetic energy due to wake effects from multiple upstream turbines.

adjoint optimization↗

DOE OSTI2020

INTEGRATION OF FLEX EQUIPMENT AND OPERATOR ACTIONS IN PLANT FORCE-ON-FORCE MODELS WITH DYNAMIC RISK ASSESSMENT

The overall operation and maintenance cost to protect nuclear power plants accounts for approximately 7% of the total cost of power generation, with labor accounting for half of this cost. In the current research, from interaction with utilities and other stakeholders, it was determined that physical security forces account for nearly 20% of the entire workforce at several nuclear power plants. Labor costs continue to rise in the U.S., so any measures to reduce the cost of operating a nuclear power plant will need to include a reduction in labor. The physical security pathway within the DOE’s Light Water Reactor Sustainability program aims to lower the cost of physical security through directed research into modeling and simulation, application of advanced sensors or deployment of advanced weapons. This report presents a modeling and simulation framework for integrating Diverse and Flexible Mitigation Capability (FLEX) portable equipment performance with Force on Force models of a plant’s physical security posture. The generic framework is described in detail, followed by a case study of modeling an adversarial attack aimed at causing a radiological release by sabotaging the plant’s power supply and its ultimate heat sink capabilities at a hypothetical nuclear power plant. Two different FLEX deployment strategies, series and parallel, are modeled with distinct timelines. The results of the adversarial attack modeled in a commercial Force on Force tool are integrated with the FLEX deployment model in INL’s dynamic modeling tool EMRALD. Monte Carlo simulation is used to model the distribution of the timeline in FLEX deployment strategies. The results demonstrate that, even in the extreme case of a successful adversarial attack, deployment of FLEX equipment can result in a significantly high likelihood of preventing radiological release. The modeling and simulation framework integrating FLEX equipment with Force on Force models enables the nuclear power plants to credit FLEX portable equipment in the plant security posture, resulting in an efficient and optimized physical security.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Scalable balanced training of conditional generative adversarial neural networks on image data

Here, we propose a distributed approach to train deep convolutional generative adversarial neural network (DC-CGANs) models. Our method reduces the imbalance between generator and discriminator by partitioning the training data according to data labels, and enhances scalability by performing a parallel training where multiple generators are concurrently trained, each one of them focusing on a single data label. Performance is assessed in terms of inception score, Fréchet inception distance, and image quality on MNIST, CIFAR10, CIFAR100, and ImageNet1k datasets, showing a significant improvement in comparison to state-of-the-art techniques to training DC-CGANs. Weak scaling is attained on all the four datasets using up to 1000 processes and 2000 NVIDIA V100 GPUs on the OLCF supercomputer Summit.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1998

Parallelization of Rocket Engine Simulator Software (PRESS)

We have outlined our work in the last half of the funding period. We have shown how a demo package for RESSAP using MPI can be done. However, we also mentioned the difficulties with the UNIX platform. We have reiterated some of the suggestions made during the presentation of the progress of the at Fourth Annual HBCU Conference. Although we have discussed, in some detail, how TURBDES/PUMPDES software can be run in parallel using MPI, at present, we are unable to experiment any further with either MPI or PVM. Due to X windows not being implemented, we are also not able to experiment further with XPVM, which it will be recalled, has a nice GUI interface. There are also some concerns, on our part, about MPI being an appropriate tool. The best thing about MPr is that it is public domain. Although and plenty of documentation exists for the intricacies of using MPI, little information is available on its actual implementations. Other than very typical, somewhat contrived examples, such as Jacobi algorithm for solving Laplace's equation, there are few examples which can readily be applied to real situations, such as in our case. In effect, the review of literature on both MPI and PVM, and there is a lot, indicate something similar to the enormous effort which was spent on LISP and LISP-like languages as tools for artificial intelligence research. During the development of a book on programming languages [12], when we searched the literature for very simple examples like taking averages, reading and writing records, multiplying matrices, etc., we could hardly find a any! Yet, so much was said and done on that topic in academic circles. It appears that we faced the same problem with MPI, where despite significant documentation, we could not find even a simple example which supports course-grain parallelism involving only a few processes. From the foregoing, it appears that a new direction may be required for more productive research during the extension period (10/19/98 - 10/18/99). At the least, the research would need to be done on Windows 95/Windows NT based platforms. Moreover, with the acquisition of Lahey Fortran package for PC platform, and the existing Borland C + + 5. 0, we can do work on C + + wrapper issues. We have carefully studied the blueprint for Space Transportation Propulsion Integrated Design Environment for the next 25 years [13] and found the inclusion of HBCUs in that effort encouraging. Especially in the long period for which a map is provided, there is no doubt that HBCUs will grow and become better equipped to do meaningful research. In the shorter period, as was suggested in our presentation at the HBCU conference, some key decisions regarding the aging Fortran based software for rocket propellants will need to be made. One important issue is whether or not object oriented languages such as C + + or Java should be used for distributed computing. Whether or not "distributed computing" is necessary for the existing software is yet another, larger, question to be tackled with.

Cezzar, Ruknet↗

DOE OSTI2025

Leveraging Pre-Built Catalogs and Object-Level Scheduling to Eliminate I/O Bottlenecks in HPC Environments

Modern High-Performance Computing (HPC) environments face mounting challenges due to the shift from large to small file datasets, along with an increasing number of users and parallelized applications. As HPC systems rely on Parallel File Systems (PFS), such as Lustre for data processing, performance bottlenecks stemming from Object Storage Target (OST) contention have become a significant concern. Existing solutions, such as LADS with its object-level scheduling approach, fall short in large-scale HPC environments due to their inability to effectively address metadata I/O bottlenecks and the growing number of I/O processes. This study highlights the pressing need for a comprehensive solution that tackles both OST contention and metadata I/O challenges in diverse HPC workloads. To address these challenges, we propose SwiftLoad, an object-level I/O scheduling framework that leverages a metadata catalog to enhance the performance and efficiency of parallel HPC utilities. The adoption of the metadata catalog mitigates the metadata I/O bottlenecks that commonly occur in HPC utilities, a challenge that is particularly pronounced in object-level I/O scheduling. SwiftLoad addresses OST contention and the uneven distribution of I/O processes across different OSTs through mathematical modeling and incorporates a Loader Configuration Module to regulate the number of I/O processes. Evaluated with two representative utilities—data deduplication profiling and data augmentation—SwiftLoad achieved performance improvements of up to 5.63x and 11.0x, respectively, on a production supercomputer.

HPC↗

NASA NTRS1997

Root-Raised Cosine Filter Implementation That Uses Canonical Signed Digits for High-Speed Digital Filter Applications

NASA Lewis Research Center's Space Communications Division has been investigating high-speed digital filters that can operate at a higher speed than those in current use for a digital modulator and demodulator (modem). Using the Canonical Signed Digits (CSD) number representation for filter coefficients is a very effective way to increase the filter's speed while reducing complexity in the digital filter hardware design. This approach is a good alternative to using an expensive parallel-processing design technique or custom, application-specific integrated circuits. Such integrated circuits may not be suitable for applications that require filter speeds faster than what application-specific integrated circuits digital signal processors can offer for a dedicated channel. When a communication channel is a dedicated, multiplication process--a costly, time-consuming process--it can be greatly simplified by a replacement of the filter coefficients with CSD numbers. A computer code written with the MATLAB software package runs the program and generates CSD-represented filter coefficients that are based on minimizing minimum mean square errors. Also, the Alta Group of Cadence's Signal Processing Workstation is used to simulate and analyze the CSD filter responses. The impulse response of the root-raised cosine filter that is used as a base model is defined. From this filter, a set of coefficients is sampled and stored in a file. For the all coefficients, the optimal CSD number for each coefficient is searched on the basis of the minimum-mean-square-errors criterion. Because the distribution of CSD numbers is not uniform, quantization errors tend to be bigger for coefficients greater than 1/2. To offset errors that occur in a region of coefficients between 1/2 to 1 and to better represent fractions with CSD numbers, an extra nonzero digit is allowed for any coefficients exceeding 1/2. This will greatly improve frequency response as well as intersymbol interference at the receiver. The frequency response of a set of collected CSD-represented filter coefficients was compared with the same filter that was conventionally implemented. Analyses show CSD-implemented filters perform as well as conventional filters. Comparison of eye diagrams and bit-error-rate curves between CSD filters and traditionally implemented filters are almost indistinguishable. However, filter complexity was reduced from almost 3.5 to 1 for CSD filters. Complete computer simulation results are available. In the near future, work will focus on building actual working digital filter hardware in a field programmable gate array (FPGA).

Kim, Heechul↗

NASA NTRS2018

Development of a Detailed Surface Chemistry Framework in DSMC

A generalized finite-rate surface chemistry framework incorporating a comprehensive list of reaction mechanisms is developed and implemented into the Direct Simulation Monte Carlo (DSMC) solver SPARTA (Stochastic PArallel Rarefied-gas Time-accurate Analyzer). The various mechanisms include adsorption, desorption, Eley-Rideal (ER), and several types of Langmuir-Hinshelwood (LH) mechanisms. The approach is to stochastically model the various competing reactions occurring on a set of active sites. Both gas-surface (e.g., adsorption, ER) and pure-surface (e.g., desorption) reaction mechanisms are incorporated, and the framework also includes catalytic or surface altering mechanisms involving the participation of the bulk-phase species (e.g., bulk carbon atoms). Marschall and MacLean developed a general formulation in which multiple phases and surface sites are used and a similar convention is adopted in the current work. Expressions for the microscopic parameters of reaction probabilities (for gas-surface reactions) and frequencies (for pure-surface reactions) that are required for DSMC are derived from the surface properties and macroscopic parameters such as rate constants, sticking coefficients, etc. The energy and angular distributions of the products are specified according to the reaction type and input parameters. This framework also presents physically consistent procedures to accurately compute the reaction probabilities and frequencies in the case of multiple reactions. The result is a modeling tool with a wide variety of surface reactions characterized via user-specified reaction rate constants, surface properties and parameters.

Surface Chemistry↗

DOE OSTI2023

Extreme-scale stochastic optimization and simulation via learning-enhanced decomposition and parallelization (Final Technical Report)

Stochastic optimization and simulation models ubiquitously arise in designing and operating complex service/engineering systems. They can be extreme in scale due to high-dimensional data and decisions, and can also involve decisions made sequentially in response to newly revealed data, both causing significant computational challenge. The objective of this research is to explore a unified framework that integrates machine learning with discrete optimization and risk-averse modeling, to improve the efficiency of decomposition paradigms for stochastic optimization and simulations at extreme scale. The models we consider represent a broad class of complex decision-making problems, where 0-1 or continuous decisions are made before and/or after knowing multiple sources of uncertainties that could be correlated. We will employ machine learning methods to dynamically decide and prioritize computational procedures, including cut generation, branching, and bounding of the optimal objective. Furthermore, the research will shed new lights on the traditional decomposition algorithms for extreme-scale computing. Deliverables of the research include new modeling and computational methods for advancing the state-of-the-art research in optimization and simulation, bringing many relevant risk-averse, data-driven optimization problems in practice within the range of tractability. Examples include distributed computing server scheduling and sensor deployment for monitoring critical infrastructures. Success in this effort will enable progress in solving multiple extreme-scale problems in the complex system design and operations arising from DoE missions in energy, environment, and national security.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2025

Using Apptainer in a Pilot-based Distributed Workload

GlideinWMS is a pilot and pressure-based workload manager for distributed scientific computing. Many experiments like CMS and Fermilab’s Neutrino experiments use it to provision elastic clusters for their analysis and simulations, split into close to a million concurrent jobs. Most user jobs require containers, and the pilots use Apptainer to set up the desired platform. For the pilots that run as regular batch jobs, Apptainer is safer, lighter, and easier to use than other containerization solutions. Many images used by the pilots are expanded SIF images distributed via the CernVM-FS: this combination is very efficient. At Fermilab, for example, we store on GitHub Dockerfiles that mimic the platform in the worker nodes of local clusters. GitHub workflows build and push the images to Docker Hub, and a service periodically pulls and converts them to the expanded SIF images in the CernVM-FS, so the scientists can find a familiar environment everywhere. Apptainer has also been used to run services inside the pilot jobs, like benchmarks that characterize the worker node being used, or a Triton Inference Server that allows sharing a GPU with all the jobs that run in parallel on a node.

Mambelli, Marco [Fermilab] (ORCID:0000000294892681↗

DOE OSTI2025

TorchBraid: High-Performance Layer-Parallel Training of Deep Neural Networks with MPI and GPU Acceleration

TorchBraid is a high-performance implementation of layer-parallel training for deep neural networks (DNNs) supporting MPI-based parallelism and GPU acceleration. Layer-parallel training has been developed to overcome the serialization inherent in forward and backward propagation of DNNs that limits utilization of computational resources in the strong scaling limit. To achieve this, TorchBraid integrates the PyTorch neural network framework with the state-of-the-art XBraid time-parallel library. Furthermore, this article presents the use and performance of TorchBraid, in addition to solutions for overcoming the algorithmic challenges inherent in combining automatic differentiation with layer-parallel. Results are presented with and without GPU acceleration for the Tiny ImageNet and MNIST image classification data sets, as well as recurrent neural networks. Overall, TorchBraid enables fast training of DNNs, both in a strong and weak scaling context. In addition to the TorchBraid software, several new advances in applying layer-parallel algorithms are detailed. Integration of layer-parallel with data-parallel algorithms is presented for the first time, showing the computational advantages of the combination. Standard deep learning techniques, like batch-normalization, are developed for layer-parallel training. Finally, a new approach combining layer-parallel with spatial coarsening in order to accelerate training for 3D image classification shows roughly a 10× speedup over serial execution.

Layer-parallel↗

DOE OSTI2024

A Full-Stack Exploration of Language-Based Parallelism in Fortran 2023

This poster explores native parallel features in Fortran 2023 through the lens of supporting applications with libraries, compilers, and parallel runtimes. The language revision informally named Fortran 2008 introduced parallelism in the form of Single Program Multiple Data (SPMD) execution with two broad feature sets: (1) loop-level parallelism via do concurrent and (2) a Partitioned Global Address Space (PGAS) comprised of distributed “coarray” data structures. Fortran’s native parallelism has demonstrated high performance [1] and reduced the burden of inserting what sometimes amounts to more directives than code. Several compilers support both feature sets, typically by translating do concurrent into serial do loops annotated by parallel directives and by translating SPMD/PGAS features into direct calls to a communication library. Our research focuses primarily on two questions: (1) can the compiler’s parallel runtime library be developed in the language being compiled (Fortran) and (2) can we define an interface to the runtime that liberates compilers from being hardwired to one runtime and vice versa. We are answering these questions by developing the Parallel Runtime Interface for Fortran (PRIF) [2] and the Co-Array Fortran Framework of Efficient Interfaces to Network Environments (Caffeine) [3]. Caffeine is initially targeting adoption by LLVM Flang, a new open-source Fortran compiler developed by a broad community in industry, academia, and government labs. We are also exploring the use of these features in Inference-Engine, a deep learning library designed to facilitate neural network training and inference for high-performance computing applications written in modern Fortran.

Rasmussen, Katherine↗