Results for “scalable computing” — Page 64

NASA NTRSDate not supplied

Rapid Spacecraft Payload Development: In-Orbit Demonstration of Flight Software Reuse, Scalability, and Dependability

As space mission design trends towards shared, multi-mission platforms and high-performance onboard computing architectures, the number of spacecraft launched into operation is also steadily rising. Through ridesharing, spacecraft miniaturization, and other cost-reduction measures, the barriers to space are lowering, resulting in compounded growth in the amount of flight software being deployed. To meet the needs of both the growing quantity and evolving nature of spacecraft, flight software design must accordingly adapt to support more efficient development, solutions to computational resource-sharing, and software reusability. This paper focuses on a software payload demonstrating several core technologies that improve the state-of-the-art in these identified areas. Launched into low-earth orbit in January 2022, our software payload was conceived, designed, and delivered in a span of merely two months. It was developed on top of the NASA core Flight System (cFS) framework and the Distributed Spacecraft Autonomy (DSA) Comm cFS application, which translates cFS software bus messages across a Data Distribution Service (DDS) network. The flight software, packaged in Linux container images, was deployed as one of 18 flight applications managed through the Unibap SpaceCloud Framework. The applications were run on a Unibap iX5-102 radiation-tolerant payload computer, hosted on the D-Orbit SCV-004 spacecraft as part of an ESA-sponsored in-orbit technology test. Our payload, referred to as the DSA D-Orbit software, demonstrates the reusability of the DSA Comm app in a substantially different context and purpose as its original mission. Comm’s original design goal was to reliably distribute messages between spacecraft swarms of arbitrary size and dynamic network topology. However, we leverage this same functionality to introduce redundancy and opportunistic parallel data processing in the context of a representative onboard image processing workload. This adaptive mission architecture was enabled in part by the SpaceCloud Framework’s use of container virtualization as the payload integration interface. By using a base container image with common high-level language runtimes and libraries, we were able to rapidly design, develop, and validate our image processing application without many of the technological barriers common to flight software development. We present details the goals, approach, results, and lessons learned through this technology demonstration experiment and contextualize those observations against present and future challenges in spacecraft software development.

computer programming↗

DOE OSTI2024

Separable physics-informed DeepONet: Breaking the curse of dimensionality in physics-informed machine learning

The deep operator network (DeepONet) has shown remarkable potential in solving partial differential equations (PDEs) by mapping between infinite-dimensional function spaces using labeled datasets. However, in scenarios lacking labeled data, the physics-informed DeepONet (PI-DeepONet) approach, which utilizes the residual loss of the governing PDE to optimize the network parameters, faces significant computational challenges, particularly due to the curse of dimensionality. This limitation has hindered its application to high-dimensional problems, making even standard 3D spatial with 1D temporal problems computationally prohibitive. Additionally, the computational requirement increases exponentially with the discretization density of the domain. Here, to address these challenges and enhance scalability for high-dimensional PDEs, we introduce the Separable physics-informed DeepONet (Sep-PI-DeepONet). This framework employs a factorization technique, utilizing sub-networks for individual one-dimensional coordinates, thereby reducing the number of forward passes and the size of the Jacobian matrix required for gradient computations. By incorporating forward-mode automatic differentiation (AD), we further optimize computational efficiency, achieving linear scaling of computational cost with discretization density and dimensionality, making our approach highly suitable for high-dimensional PDEs. We demonstrate the effectiveness of Sep-PI-DeepONet through three benchmark PDE models: the viscous Burgers’ equation, Biot’s consolidation theory, and a parameterized heat equation. Our framework maintains accuracy comparable to the conventional PI-DeepONet while reducing training time by two orders of magnitude. Notably, for the heat equation solved as a 4D problem, the conventional PI-DeepONet was computationally infeasible (estimated 289.35 h), while the Sep-PI-DeepONet completed training in just 2.5 h. These results underscore the potential of Sep-PI-DeepONet in efficiently solving complex, high-dimensional PDEs, marking a significant advancement in physics-informed machine learning.

Neural operator↗

DOE OSTI2016

A Distributed OpenCL Framework using Redundant Computation and Data Replication

Applications written solely in OpenCL or CUDA cannot execute on a cluster as a whole. Most previous approaches that extend these programming models to clusters are based on a common idea: designating a centralized host node and coordinating the other nodes with the host for computation. However, the centralized host node is a serious performance bottleneck when the number of nodes is large. In this paper, we propose a scalable and distributed OpenCL framework called SnuCL-D for large-scale clusters. SnuCL-D's remote device virtualization provides an OpenCL application with an illusion that all compute devices in a cluster are confined in a single node. To reduce the amount of control-message and data communication between nodes, SnuCL-D replicates the OpenCL host program execution and data in each node. We also propose a new OpenCL host API function and a queueing optimization technique that significantly reduce the overhead incurred by the previous centralized approaches. To show the effectiveness of SnuCL-D, we evaluate SnuCL-D with a microbenchmark and eleven benchmark applications on a large-scale CPU cluster and a medium-scale GPU cluster.

Kim, Junghyun↗

DOE OSTI2020

NRC Multiphysics Analysis Capability Deployment FY2020: Part 3

This report details progress and activities of Idaho National Laboratory (INL) on the Nuclear Regulatory Commission (NRC) project “Development and Modeling Sup- port for Advanced Non-Light Water Reactors.” The tasks completed for this report are: Task2c: Explicit modeling of pebble transient temperature response. In this simulation, the 400 MWth Pebble-Bed Modular Reactor (PBMR) design, PBMR- 400, experiences a 20-second power ramp from 100% to 150% power. This is followed by a similar reduction in the power back to 100%. Several multiscale pebble coupling approaches are tested with one pebble per mesh element in the active core region. The results show good conservation behavior and the stability of the coupling.; Extended scope part 1: An assessment of the computational efficiency of the Discontinuous Finite Element Method (DFEM) heat transfer solver shows good scalability. The DFEM solver is a factor of 4 more expensive in solution time than the Finite Element Method (FEM) solver for heat transfer problems due to the increased number of degrees of freedom. Nonetheless, the DFEM approach provides the user with the flexibility to model gap heat transfer problems.; Extended scope part 2: The GapHeatTransferInterfaceMaterial was improved to give the user increased flexibility with the modeling of heat transfer through gaps with the DFEM solver. A number of gap parameters can now be coupled both through functions and variables.; Extended scope part 3: Demonstration of how the gap width between hexagonal fuel cells can be calculated during a heat-up transient and used in the GapHeatTransferInterface model. A full-domain DFEM model with gap expansion is coupled to a SubApp that models the thermal expansion of the base plate. The results show the expected physical behavior, although have not been fully bench-marked at this point in time.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2024

Enhancing Distribution System Resilience: A First-Order Meta-RL Algorithm for Critical Load Restoration

The increasing frequency of extreme events and the integration of distributed energy resources (DERs) into modern grids have elevated the need for resilient and efficient critical load restoration strategies in distribution systems. However, the stochastic nature of renewable DERs, limited energy resource availability and the intricate nonlinearities inherent in complex grid control problem make the problem challenging. Although reinforcement learning (RL) and warm-start RL methods have shown promising results, their performance often falls short in rapidly adapting to new, unseen situations and typically requires exhaustive problem-specific tuning. To address these gaps, we propose a First-Order Meta-based RL (FOM-RL) algorithm within an online framework for adaptive and robust critical load restoration. By harnessing local DERs as the enabling technology, FOM-RL allows the RL agent to swiftly adapt to new unseen scenarios by leveraging previously acquired knowledge of different tasks. Experimental results provide evidence that proposed algorithm learns more efficiently and showcases generalization capabilities across diverse set of operational scenarios. Moreover, a rigorous theoretical analysis yields a tight sublinear regret bound, sensitive to temporal variability, with a task-averaged optimality gap bounded by O(VM+D*/(Tsquare root(M))). These results suggest that optimality improves with task similarity and an increased number of tasks M, reaffirming the efficacy and scalability of the proposed approach in addressing the complexities of critical load restoration in distribution systems.

complexity theory↗

DOE OSTI2020

Scalable Asynchronous Domain Decomposition Solvers

We discuss how parallel implementations of linear iterative solvers generally alternate between phases of data exchange and phases of local computation. Increasingly large problem sizes and more heterogeneous compute architectures make load balancing and the design of low latency network interconnects that are able to satisfy the communication requirements of linear solvers very challenging tasks. In particular, global communication patterns such as inner products become increasingly limiting at scale. We explore the use of asynchronous communication based on one-sided Message Passing Interface primitives in the context of domain decomposition solvers. In particular, a scalable asynchronous two-level Schwarz method is presented. We discuss practical issues encountered in the development of a scalable solver and show experimental results obtained on a state-of-the-art supercomputer system that illustrate the benefits of asynchronous solvers in load balanced as well as load imbalanced scenarios. Using the novel method, we can observe speedups of up to four times over its classical synchronous equivalent.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2015

Accelerating k-NN Algorithm with Hybrid MPI and OpenSHMEM

Machine Learning algorithms are benefiting from the continuous improvement of programming models, including MPI, MapReduce and PGAS. k-Nearest Neighbors (k-NN) algorithm is a widely used machine learning algorithm, applied to supervised learning tasks such as classification. Several parallel implementations of k-NN have been proposed in the literature and practice. However, on high-performance computing systems with high-speed interconnects, it is important to further accelerate existing designs of the k-NN algorithm through taking advantage of scalable programming models. To improve the performance of k-NN on large-scale environment with InfiniBand network, this paper proposes several alternative hybrid MPI+OpenSHMEM designs and performs a systemic evaluation and analysis on typical workloads. The hybrid designs leverage the one-sided memory access to better overlap communication with computation than the existing pure MPI design, and propose better schemes for efficient buffer management. The implementation based on k-NN program from MaTEx with MVAPICH2-X (Unified MPI+PGAS Communication Runtime over InfiniBand) shows up to 9.0% time reduction for training KDD Cup 2010 workload over 512 cores, and 27.6% time reduction for small workload with balanced communication and computation. Experiments of running with varied number of cores show that our design can maintain good scalability.

Lin, Jian↗

DOE OSTI2017

Extreme Scale Unstructured Adaptive CFD: From Multiphase Flow to Aerodynamic Flow Control

Understanding the flow of fluid, either liquid or gas, through and around solid bodies has challenged man since the dawn of scientific inquiry. Many of the great minds of science and math have progressively built up a hierarchy of fluid models. This report is concerned with the computational modeling of turbulent flow around aerodynamic bodies such as planes and wind turbines. In this case, viscous effects near the solid bodies create very thin boundary layers that yield highly anisotropic (gradients normal to the surface may be 10^6 times larger than gradients along the surface) solutions to the governing non-linear partial differential equations (PDE); the Navier-Stokes equations. Furthermore, turbulent flows develop extremely broad ranges of length and time scales.This disparity motivates the use of discretization methods capable of employing adaptivity and implicit time integration. The combination of these features (non-linear, anisotropy, adaptivity, andimplicit) dramatically raise the complexity of the discretization, posing large challenges to efficient scalable parallel implementation. However, through careful design, the more complex algorithms can provide great reductions in computational cost relative to simpler methods (e.g., Cartesian grids with explicit time integration) that are easier to mate efficiently to hardware. In this report, we not only describe our approach but we also address the fact that while complex algorithms may never be as efficient flop-for-flop as simple methods, in the important measure of science-per-core-hour, they can still win big by making complex features like adaptivity and implicit methods as efficient and scalable as possible.

42 ENGINEERING↗

DOE OSTI2025

Hybrid learning techniques for scientific data reduction with performance guarantees

The research initiatives supported by the U.S. Department of Energy (DOE) Grant DE-SC0022265 are fundamentally aimed at pioneering advanced machine learning (ML) techniques for scientific data compression within high-performance computing (HPC) environments. This comprehensive body of work addresses the critical challenge posed by the exponential growth of data generated by scientific simulations in domains such as fusion energy, climate modeling, and computational fluid dynamics (CFD). A core objective is to develop compression algorithms that achieve substantial data reduction—often by orders of magnitude—while rigorously ensuring the fidelity of both the primary data (PD) and scientifically crucial derived quantities of interest (QoI). The methodologies deployed under this grant integrate sophisticated deep learning architectures, prominently featuring autoencoders, advanced generative models like conditional diffusion, and hybrid learning techniques. Key innovations include the development of Guaranteed Autoencoders (GAE) and the Guaranteed Conditional Diffusion with Tensor Correction (GCDTC) framework, which provide explicit, instance-level error bounds on reconstructed data. Furthermore, specialized strategies such as nonlinear constraint satisfaction are employed to preserve the integrity of QoI, a vital requirement for the trustworthiness of downstream scientific analyses. This research also focuses on the design and implementation of scalable, GPU-accelerated software pipelines that seamlessly integrate into existing HPC workflows, ensuring both computational efficiency and practical applicability. The CAESAR framework, for example, unifies foundation and generative models to create an adaptive and efficient compression solution for spatio-temporal scientific data. Collectively, these efforts represent a significant advancement in mitigating the scientific data deluge, enabling more effective data management, accelerated scientific discovery, and optimized utilization of HPC resources.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Final report- UFL - RAPIDS2: A SciDAC Institute for Computer Science, Data, and Artificial Intelligence

The research initiatives supported by the U.S. Department of Energy (DOE) Grant DE-SC0022265 are fundamentally aimed at pioneering advanced machine learning (ML) techniques for scientific data compression within high-performance computing (HPC) environments. This comprehensive body of work addresses the critical challenge posed by the exponential growth of data generated by scientific simulations in domains such as fusion energy, climate modeling, and computational fluid dynamics (CFD). A core objective is to develop compression algorithms that achieve substantial data reduction—often by orders of magnitude—while rigorously ensuring the fidelity of both the primary data (PD) and scientifically crucial derived quantities of interest (QoI). The methodologies deployed under this grant integrate sophisticated deep learning architectures, prominently featuring autoencoders, advanced generative models like conditional diffusion, and hybrid learning techniques. Key innovations include the development of Guaranteed Autoencoders (GAE) and the Guaranteed Conditional Diffusion with Tensor Correction (GCDTC) framework, which provide explicit, instance-level error bounds on reconstructed data. Furthermore, specialized strategies such as nonlinear constraint satisfaction are employed to preserve the integrity of QoI, a vital requirement for the trustworthiness of downstream scientific analyses. This research also focuses on the design and implementation of scalable, GPU-accelerated software pipelines that seamlessly integrate into existing HPC workflows, ensuring both computational efficiency and practical applicability. The CAESAR framework, for example, unifies foundation and generative models to create an adaptive and efficient compression solution for spatio-temporal scientific data. Collectively, these efforts represent a significant advancement in mitigating the scientific data deluge, enabling more effective data management, accelerated scientific discovery, and optimized utilization of HPC resources.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Reconfigurable training and reservoir computing in an artificial spin-vortex ice via spin-wave fingerprinting

Strongly interacting artificial spin systems are moving beyond mimicking naturally occurring materials to emerge as versatile functional platforms, from reconfigurable magnonics to neuromorphic computing. Typically, artificial spin systems comprise nanomagnets with a single magnetization texture: collinear macrospins or chiral vortices. Here, by tuning nanoarray dimensions we have achieved macrospin–vortex bistability and demonstrated a four-state metamaterial spin system, the ‘artificial spin-vortex ice’ (ASVI). ASVI can host Ising-like macrospins with strong ice-like vertex interactions and weakly coupled vortices with low stray dipolar field. Vortices and macrospins exhibit starkly differing spin-wave spectra with analogue mode amplitude control and mode frequency shifts of Δf = 3.8 GHz. The enhanced bitextural microstate space gives rise to emergent physical memory phenomena, with ratchet-like vortex injection and history-dependent non-linear fading memory when driven through global magnetic field cycles. We employed spin-wave microstate fingerprinting for rapid, scalable readout of vortex and macrospin populations, and leveraged this for spin-wave reservoir computation. ASVI performs non-linear mapping transformations of diverse input and target signals in addition to chaotic time-series forecasting.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Surrogate Model Integration with MOOSE XFEM for Creep Crack Growth

Ferritic-martensitic steels are key structural materials for advanced reactors but experience time-dependent deformation and damage under prolonged high temperature and irradiation, leading to creep-driven crack initiation and growth. High-fidelity models—crystal plasticity with irradiation mechanisms, phase-field for microstructural evolution, and continuum-damage viscoplasticity—capture the underlying physics but are too computationally intensive for broad design-space exploration and uncertainty quantification. This milestone advances a scalable alternative by integrating a microstructure-sensitive surrogate creep model into the Multiphysics Object-Oriented Simulation Environment (MOOSE) finite element framework and extending it to fracture via the extended finite element method (XFEM). The surrogate model, developed with collaborators at Sandia and Los Alamos National Laboratories, maps relevant microstructural descriptors to the viscoplastic response of HT9. We embed this surrogate within a coupled deformation-damage workflow in MOOSE/XFEM to simulate creep-driven crack initiation and propagation. Implementation enhancements include updates to the material interface, a plastic correction phase involving microstructure evolution, and fracture criteria to ensure numerical robustness and compatibility with the surrogate structure. Demonstrations on canonical creep benchmarks spanning uniaxial and multiaxial states show that the surrogate reproduces key trends of high-fidelity models while substantially reducing computational cost. The resulting capability bridges physics fidelity and performance, providing a practical path to a predictive, microstructure-aware assessment of creep and fracture in reactor materials.

36 - MATERIALS SCIENCE↗

DOE OSTI2017

Building Energy Management Open Source Software

This is the repository for Building Energy Management Open Source Software (BEMOSS), which is an open source operating system that is engineered to improve sensing and control of equipment in small- and medium-sized commercial buildings. BEMOSS offers the following key features: (1) Open source, open architecture – BEMOSS is an open source operating system that is built upon VOLTTRON – a distributed agent platform developed by Pacific Northwest National Laboratory (PNNL). BEMOSS was designed to make it easy for hardware manufacturers to seamlessly interface their devices with BEMOSS. Software developers can also contribute to adding additional BEMOSS functionalities and applications. (2) Plug & play – BEMOSS was designed to automatically discover supported load controllers (including smart thermostats, VAV/RTUs, lighting load controllers and plug load controllers) in commercial buildings. (3) Interoperability – BEMOSS was designed to work with load control devices form different manufacturers that operate on different communication technologies and data exchange protocols. (4) Cost effectiveness – Implementation of BEMOSS deemed to be cost-effective as it was built upon a robust open source platform that can operate on a low-cost single-board computer, such as Odroid. This feature could contribute to its rapid deployment in small- or medium-sized commercial buildings. (5) Scalability and ease of deployment – With its multi-node architecture, BEMOSS provides a distributed architecture where load controllers in a multi-floor and high occupancy building could be monitored and controlled by multiple single-board computers hosting BEMOSS. This makes it possible for a building engineer to deploy BEMOSS in one zone of a building, be comfortable with its operation, and later on expand the deployment to the entire building to make it more energy efficient. (6) Ability to provide local and remote monitoring – BEMOSS provides both local and remote monitoring ability with role-based access control. (7) Security – In addition to built-in security features provided by VOLTTRON, BEMOSS provides enhanced security features, including BEMOSS discovery approval process, encrypted core-to-node communication, thermostat anti-tampering feature and many more. (8) Support from the Advisory Committee – BEMOSS was developed in consultation with an advisory committee from the beginning of the project. BEMOSS advisory committee comprises representatives from 22 organizations from government and industry.

Source record↗

DOE OSTI2017

Building Energy Management Open Source Software

This is the repository for Building Energy Management Open Source Software (BEMOSS), which is an open source operating system that is engineered to improve sensing and control of equipment in small- and medium-sized commercial buildings. BEMOSS offers the following key features: (1) Open source, open architecture – BEMOSS is an open source operating system that is built upon VOLTTRON – a distributed agent platform developed by Pacific Northwest National Laboratory (PNNL). BEMOSS was designed to make it easy for hardware manufacturers to seamlessly interface their devices with BEMOSS. Software developers can also contribute to adding additional BEMOSS functionalities and applications. (2) Plug & play – BEMOSS was designed to automatically discover supported load controllers (including smart thermostats, VAV/RTUs, lighting load controllers and plug load controllers) in commercial buildings. (3) Interoperability – BEMOSS was designed to work with load control devices form different manufacturers that operate on different communication technologies and data exchange protocols. (4) Cost effectiveness – Implementation of BEMOSS deemed to be cost-effective as it was built upon a robust open source platform that can operate on a low-cost single-board computer, such as Odroid. This feature could contribute to its rapid deployment in small- or medium-sized commercial buildings. (5) Scalability and ease of deployment – With its multi-node architecture, BEMOSS provides a distributed architecture where load controllers in a multi-floor and high occupancy building could be monitored and controlled by multiple single-board computers hosting BEMOSS. This makes it possible for a building engineer to deploy BEMOSS in one zone of a building, be comfortable with its operation, and later on expand the deployment to the entire building to make it more energy efficient. (6) Ability to provide local and remote monitoring – BEMOSS provides both local and remote monitoring ability with role-based access control. (7) Security – In addition to built-in security features provided by VOLTTRON, BEMOSS provides enhanced security features, including BEMOSS discovery approval process, encrypted core-to-node communication, thermostat anti-tampering feature and many more. (8) Support from the Advisory Committee – BEMOSS was developed in consultation with an advisory committee from the beginning of the project. BEMOSS advisory committee comprises representatives from 22 organizations from government and industry.

Source record↗

DOE OSTI2017

A Multilevel, Hierarchical Sampling Technique for Spatially Correlated Random Fields

In this paper, we propose an alternative method to generate samples of a spatially correlated random field with applications to large-scale problems for forward propagation of uncertainty. A classical approach for generating these samples is the Karhunen--Loève (KL) decomposition. However, the KL expansion requires solving a dense eigenvalue problem and is therefore computationally infeasible for large-scale problems. Sampling methods based on stochastic partial differential equations provide a highly scalable way to sample Gaussian fields, but the resulting parametrization is mesh dependent. We propose a multilevel decomposition of the stochastic field to allow for scalable, hierarchical sampling based on solving a mixed finite element formulation of a stochastic reaction-diffusion equation with a random, white noise source function. Lastly, numerical experiments are presented to demonstrate the scalability of the sampling method as well as numerical results of multilevel Monte Carlo simulations for a subsurface porous media flow application using the proposed sampling method.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Network Anomaly Detection Using Federated Learning

The internet is turning out to be an integral part of every-one's lives as more and more devices are being connected to serve societal needs. Our work is motivated by two ma-jor observations. Firstly, one drawback of connecting to the network is the threat of network attacks that can compromise users' private information, leading to data loss and adversely affecting productivity. There are several traditional security mechanisms to defend against these attacks, such as firewalls, virtual private networks (VPNs), demilitarized zones (DMZs), and vulnerability scanners. One way to prevent these attacks is early detection and prevention. However, these kinds of architecture do not scale very well because of their centralized nature. Secondly, we observe from heuristics and data set distributions that the majority of the requests made to a server are innocuous. Therefore, almost all server request data sets are highly imbalanced, weighted highly towards the harmless requests.

Marfo, William↗

DOE OSTI2016

A Bloom Filter Based Scalable Data Integrity Check Tool for Large-Scale Dataset

Large scale HPC applications are becoming increasingly data intensive. At Oak Ridge Leadership Computing Facility (OLCF), we are observing the number of files curated under individual project are reaching as high as 200 millions and project data size is exceeding petabytes. These simulation datasets, once validated, often needs to be transferred to archival system for long term storage or shared with the rest of the research community. Ensuring the data integrity of the full dataset at this scale is paramount important but also a daunting task. This is especially true considering that most conventional tools are serial and file-based, unwieldy to use and/or can't scale to meet user's demand.To tackle this particular challenge, this paper presents the design, implementation and evaluation of a scalable parallel checksumming tool, fsum, which we developed at OLCF. It is built upon the principle of parallel tree walk and work-stealing pattern to maximize parallelism and is capable of generating a single, consistent signature for the entire dataset at extreme scale. We also applied a novel bloom-filter based technique in aggregating signatures to overcome the signature ordering requirement. Given the probabilistic nature of bloom filter, we provided a detailed error and trade-off analysis. Using multiple datasets from production environment, we demonstrated that our tool can efficiently handle both very large files as well as many small-file based datasets. Our preliminary test showed that on the same hardware, it outperforms conventional tool by as much as 4×. It also exhibited near-linear scaling properties when provisioned with more compute resources.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2017

Improving Uintah's Scalability Through the Use of Portable Kokkos-Based Data Parallel Tasks

The University of Utah’s Carbon Capture Multidisciplinary Simulation Center (CCMSC) is using the Uintah Computational Framework to predict performance of a 1000 MWe ultra-supercritical clean coal boiler. The center aims to utilize the Intel Xeon Phi-based DOE systems, Theta and Aurora, through the Aurora Early Science Program by using the Kokkos C++ library to enable node-level performance portability. This paper describes infrastructure advancements and portability improvements made possible by the integration of Kokkos within Uintah. This integration marks a step towards consolidating Uintah’s MPI+PThreads and MPI+CUDA hybrid parallelism approaches into a single MPI+Kokkos approach. Scalability results are presented that compare serial and data parallel task execution models for a challenging radiative heat transfer calculation, central to the center’s predictive boiler simulations. Here, these results demonstrate both good strong-scaling characteristics to 256 Knights Landing (KNL) processors on the NSF Stampede system, and show the KNL-based calculation to compete with prior GPU-based results for the same calculation.

97 MATHEMATICS AND COMPUTING↗