Results for “programming model” — Page 55

DOE OSTI2025

Meso-scale modeling of UO2 nuclear fuel to high burnup

To improve the economics of light water reactors for commercial nuclear energy generation, utility operators are seeking to obtain regulatory approval to run UO2 fuel to higher levels of burnup. One potential impediment to obtaining this approval is the phenomenon of fuel fragmentation, relocation, and dispersal (FFRD). FFRD can result when fuel experiences a rapid temperature transient, such as that occurring during a Loss Of Coolant Accident (LOCA). FFRD has historically been most associated with the rim region in UO2 fuel pellets, where the phenomenon of fragmentation is also referred to as pulverization due to the small size of the fragments. More recent evidence suggests that the so-called “dark zone” (due to its appearance in micrographs) that can be observed in the mid-radial regions of high burnup fuel is also susceptible to FFRD. Although empirical fuel performance models have been developed that can adequately predict pulverization in the rim region under typical LWR conditions, a scientific understanding of what underlies fuel restructuring and subsequent FFRD is lacking even in the rim region, and no models are currently available for the behavior the dark zone. To address these challenges, the U.S. Department of Energy’s Nuclear Energy Advanced Modeling and Simulation (NEAMS) program has employed a multi-scale modeling approach to improve scientific understanding and develop new fuel performance models. In this talk, I will focus on meso-scale efforts, which form a crucial link between atomic-scale and engineering-scale models. Phase-field modeling combined with cluster dynamics is used to predict the restructuring process in the rim region. Phase-field fracture modeling, informed by atomistic simulations, is used to predict the onset of pulverization in the rim region. Combining these techniques together allows the extent of rim pulverization to be predicted. The formation and evolution of the dark zone has also been simulated with the phase-field method, using an improved approach to vacancy source term parameterization. The work shows the important impact of microstructure on fuel performance.

fracture↗

DOE OSTI2018

Aqueous Degradation in Harvested Medium Voltage Cables in Nuclear Power Plants

For medium voltage (MV) power cables with voltages between 5 and 35 kV that provide power to emergency and safety support systems in nuclear power plants (NPPs), degradation and failure of cables due to water exposure has occurred. Systematic studies of current NPP MV cable systems are being carried out to determine and characterize cable-aging mechanisms and provide greater accuracy to models being developed to predict MV cable field performance to support existing cable aging management and monitoring programs. To validate models based on representative laboratory specimens, samples from harvested MV cable systems are being subjected to accelerated aging and degradation from humid conditions and submergence. Degradation will be characterized via partial discharge and voltage endurance testing of the cable and induced water tree growth in insulation. The technical approach to be used for the testing harvested MV cable samples is presented.

Duckworth, Robert↗

DOE OSTI2016

Verified! How Performance-Based Procurement Can Boost New Construction Program Savings and Empower Building Owners

Most utility new construction programs depend on modeled projections and deemed savings based on individual technologies. The industry has established ways to evaluate these programs based on the promise - not the actual performance - of these buildings post- occupancy. Deeper savings from a more integrated design with continuous performance monitoring are left on the table, locking out the higher performance for the lifetime of the building. This paper will demonstrate how performance-based procurement can empower utility incentive offerings to drive deep energy savings and reduce administrative costs. The key is to incentivize a change in the building owner's procurement process and reward actual building energy performance rather than incentivizing specific measures. The paper looks at strategies to transition new construction programs from technology-based verification to actual measurement of performance outcomes. Accelerate Performance, a U.S. Department of Energy (DOE) funded initiative, changes the building procurement approach, which spurs innovation and motivates the design and contracting industry to achieve deep energy savings in new construction - in the case of DOE's National Renewable Energy Laboratory Research Support Facility (NREL) in Golden, CO, 50% without increasing initial costs. Utility programs in three states, Illinois, Minnesota, and Connecticut, are actively piloting the approach with leading-edge owners. The paper will share savings results and lessons learned from the three different utility pilot programs. The paper provides practical guidance for utility program managers about how to apply the Accelerate Performance approach to supplement their current new construction program offerings.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2025

Business Models for Scaling Demand Flexibility Volume IV – Program life cycle challenges and lessons learned from U.S. programs

Load growth at the grid edge is driving increased attention to the distribution system and its ability to enable customer technology adoption in an affordable and timely manner. Key industry stakeholders, including electric utilities and regulators, can benefit from strategies to manage and balance customer needs with infrastructure investments, such as demand flexibility. This report focuses on demand flexibility—the ability to reduce, shift, shed, generate, or modulate loads in response to building and grid needs—to reduce the need for costly grid upgrades by deferring investment needs and increase system reliability by shifting electricity usage during periods of high risk. Specifically, we focus on the emerging characteristics of business models for demand flexibility as a framework to understand how demand flexibility programs generate value. In this report, we focus on the life cycle of demand flexibility programs, which provides information on value creation and describes the various deployment phases program implementers navigate from initial program conceptualization through to program expansion and replication to new customer segments and regions. This report characterizes the key phases of the demand flexibility program life cycle, identifies existing challenges across the program deployment phases, and describes lessons learned. This report is part of a series that includes reports on customer relationship management strategies, stakeholder ecosystem management, and program life cycle.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2025

Updating and Refining BISON TRISO Assessment

Several advanced reactor concepts consider tristructural isotropic (TRISO) fuel in their designs. To support the advanced reactor industry, the Nuclear Energy Advanced Modeling and Simulation (NEAMS) program within the U.S. Department of Energy (DOE) is developing modeling and simulation capabilities within the BISON fuel performance code. Assessment of these capabilities against available experi- mental data, such as the Advanced Gas Reactor (AGR) program, is essential in instilling confidence in the code predictions. Associated uncertainties in model inadequacy, experimental measurements, and manufacturing are also of importance. This report discusses the updating and standardization of the existing assessment suite in BISON for TRISO fuel performance analysis. The cases considered the AGR-1, AGR-2, and AGR-3/4 experimental programs. The report also includes a preliminary investigation into the sources of model inadequacy associated with silver (Ag) release fractions using the AGR-2 Compact 2-1-2 as a case study. The results indicate that uncertainties in the operating temperatures supplied to the model and the form of those temperature conditions (spatially averaged versus spatially resolved) both affect Ag release predictions. This suggests that previously observed model inadequacy can be attributed to both physical models and the operating conditions supplied to them. These findings will be leveraged to guide development of assessment cases for additional experiments and refinement of physical models for behaviors that potentially influence Ag release, such as palladium penetration.

11 - NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

DOE OSTI2025

Towards An Updated Simulation of the Booster Neutrino Beam

Precise, accurate neutrino flux predictions for neutrino beam experiments are crucial for physics results. Flux predictions for the Booster Neutrino Beam at Fermilab were first published in 2009 by the MiniBooNE collaboration. It is no longer possible to run the simulation used to create these predictions due to outdated software versions. In this exciting period for the Short-Baseline Neutrino Program at Fermilab, with both the far detector (ICARUS) and the near detector (SBND) operating, an updated flux model for the BNB is necessary. For the purpose of using a dynamic, up-to-date beam model, the SBN program has created a new simulation named G4BNB. This framework contains new features, such as a full neutrino ancestry to handle hadron production systematics with more precision, and the inclusion of neutral mesons from the proton-beryllium scatter to allow the study of exotic BSM scenarios.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2023

Compare predictions of transient fission gas release by empirical and mechanistic models to experiments in high burnup UO 2 fuel

Understanding and predicting fuel performance at high burnup require improving our understanding of transient fission gas release. High-burnup operations enable new mechanisms of fission gas release, which affect fuel performance. The Nuclear Regulatory Commission has recently published its interpretation of existing fuel fragmentation, relocation, and dispersal data in a research information letter. There, transient fission gas release was identified as one of the main factors that contributes to fuel fragmentation, relocation, and dispersal, and therefore limits fuel extension to high burnup. However, transient fission gas release is a complex phenomenon that cannot be fully described by simple empirical descriptions. This report summarizes the development of a mechanistic model for high-burnup transient fission gas release in the fuel performance code BISON. This research was supported by the Nuclear Energy Advanced Modeling and Simulation program during fiscal year 2023 to improve our understanding of high-burnup transient fission gas release and ability to predict it as a function of operation history. To support the development of a mechanistic transient fission gas release model, the existing Simple Integrated Fission Gas Release and Swelling (Sifgrs) model in BISON has been completely refactored to make it more modular and extensible. This effort supports the model's application to high-burnup conditions, its extension to other fuel forms, and the continuous improvement of its current features. Once refactoring was completed, models for high-burnup structure formation, fission gas transfer from non-restructured fuel to high-burnup structure, high-burnup structure intragranular and intergranular fission gas behavior, high-burnup structure bubble evolution, fuel pulverization, and the resulting transient fission gas release were tested and implemented in the Simple Integrated Fission Gas Release and Swelling (Sifgrs) model or tightly coupled to it. The new mechanistic model was then compared to an empirical model developed in parallel by a Nuclear Energy University Program project using a Studsvik high-burnup loss-of-coolant-accident assessment case. Finally, the report details the preliminary BISON results for a benchmark activity organized by the Nuclear Energy Agency to evaluate fuel performance codes' predictive capabilities for burst fission gas release. This work represents an important step toward a mechanistic understanding of fission gas release in high-burnup conditions.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

DOE OSTI2016

Summary of the Persistent Current Effect Measurements in Nb 3 Sn and NbTi Accelerator Magnets at Fermilab

In the past 10 years, Fermilab has been executing an intensive R&D program on accelerator magnets based on Nb 3 Sn superconductor technology. This R&D effort includes dipole and quadrupole models for different programs, such as LARP and 11 T dipoles for the LHC high-luminosity upgrade. Before the Nb 3 Sn R&D program, Fermilab was involved in the production of the low-beta quadrupole magnets for LHC based on the NbTi superconductor. Additionally, during the 2003-2005 campaign to optimize the operation of the Tevatron, a large number of Tevatron magnets were re-measured. As a result of this field analysis, a systematic study of the persistent current decay and snapback effect in these magnets was performed. This paper summarizes the result of this study and presents a comparison between Nb 3 Sn and NbTi dipoles and quadrupoles.

43 PARTICLE ACCELERATORS↗

DOE OSTI2025

High-Burnup BWR LOCA Burst Analysis Using High-Fidelity Multiphysics Simulations

The US nuclear industry is looking to improve on the operating economics of the current fleet of light-water reactors (LWRs). One way of achieving this is by operating fuel to higher burnup. In pressurized water reactors (PWRs), relaxing the current burnup limit will allow for cycle length extensions and power uprates; in boiling water reactors (BWRs) it may allow for improved fuel utilization and reduced feed assemblies, as well as more efficient power uprates and increased capacity factors that will support the Administration’s Executive Order to facilitate 5 GW of power uprates at existing nuclear facilities. However, one of the key limitations to operating fuel to higher burnup is the risk of fuel fragmentation, relocation, and dispersal (FFRD). Recognizing the high interest in extending burnup limits, the US Nuclear Regulatory Commission (NRC) has issued Draft Regulatory Guide DG-1434, which defines an approach that would be acceptable to the NRC for addressing FFRD risk. The approach defined will require better understanding of the phenomena leading to FFRD as well as best-estimate simulation methods to understand FFRD risk in high-burnup cores. The Nuclear Energy Advanced Modeling and Simulation program is supporting the FFRD industry challenge problem through development of state-of-the-art, high-fidelity modeling and simulation LWR analysis capabilities; namely, the BISON fuel performance code and the VERA core simulator software. These tools, along with the US NRC TRACE system analysis code, have been utilized for analysis of FFRD risk in both PWR and BWR cores in recent years. The work documented in this report addresses the lack of high-fidelity research for BWRs and builds on a previous activity where the framework has been applied to Cycles 16 through 18 of Limerick Unit 1, a BWR/4, with introduction of 8 high-burnup lead use assemblies (HBLUAs) that were representative of the 8 HBLUAs loaded into Limerick Unit 2 in 2021. VERA was used in this previous activity to model rod-by-rod depletion in these cycles, and its solution was used to initialize a TRACE simulation of a large-break loss-of-coolant accident (LBLOCA) at the end of Cycle 18. In the work documented in this report, the TRACE model was improved by refining the core mesh and utilizing a new feature that allows for capturing the full 3D VERA power distribution in the model. This allows for a more detailed solution for setting BISON boundary conditions. Furthermore, the solutions from VERA and TRACE were used to set up and perform BISON simulations of about 1,000 rods sampled from the core, including all burnup levels. Utilizing two cladding burst models, it was shown that no fuel rods were predicted to burst during the postulated LBLOCA transient. Additionally, a sensitivity study was performed by artificially increasing linear heat rate during the postulated LBLOCA to identify parameters that correlate with rod burst susceptibility. Burnup, fission gas release, and hoop strain were all found to be positively correlated with rod burst susceptibility. Small-break loss-of-coolant accident (SBLOCA) analyses were also performed; these analyses predicted cladding temperature increases that were bounded by the LBLOCA cladding temperatures for all small break sizes studied for this plant. However, future refinements to the plant response assumptions during the SBLOCA could impact the predicted cladding response. Finally, a benchmark study was performed between CTF and TRACE for LOCA conditions to better qualify CTF for BWR LOCA modeling.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2016

A High-Fidelity Approach for the Simulation of Flow-Induced Vibration

Flow-induced vibration (FIV) is a widespread problem in energy systems becausethey rely on fluid movement for energy conversion. Vibrating structures may be damaged as fatigue or wear occur. Given the importance of reliable components in the nuclear industry, flow-induced vibrations have long been a major concern in the safety and operation of nuclear reactors. In particular,nuclear fuel and steam generators have been known to suffer from flow-induced vibrations and related failures. Over the past five years, the Nuclear EnergyAdvanced Modeling and Simulation program has developed the integrated multiphysics code suite SHARP. The goal of developing such a tool is to perform multiphysics modeling of the components inside a reactor core, the full reactor core or portions of it, and be able to achieve that with various levelsof fidelity. This flexibility allows users to select the appropriate level of fidelity for their computational resources and design constraints.In particular SHARP contains high-fidelity single-physics codes for structural mechanics and fluid mechanics calculations: the structural mechanics implicit code Diablo and the computational fluid dynamics spectral element code Nek5000. Both codes are state-of-the-art.highly scalable (up to millions of processors in the case of Nek5000) tools that have been extensively validated. These tools form a strong basis on which to build an FIV modeling capability.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

DOE OSTI2018

Capacity Optimization of a Centralized Charging Station in Joint Operation with a Wind Farm

In the context of large-scale wind power integration and rapid development of electric vehicles (EVs), a joint operation pattern was proposed to use a centralized charging station (CCS) to address high uncertainties incurred by wind power integration. Additionally, this would directly remove the significant indeterminacy of wind power. Because the CCS is adjacent to a wind power gathering station, it could work jointly with wind farms to operate in power system as an independent enterprise. By combining the actual operational characteristics of the wind farm and the CCS, a multidimensional operating index evaluation system was created for the joint system. Based on the wind farm’s known capacity, a multi-target capacity planning model was established for CCS to maximize the probability of realizing diverse indices of the system based on dependent-chance goal programming. In this model, planning and dispatching are combined to improve the feasibility of results in the operational stage. This approach takes the effects of randomness into account for wind power and battery swapping. The model was solved through combining Monte Carlo simulation and a genetic algorithm (GA) based on segmented encoding. Finally, as shown in the simulation results, the proposed model could comprehensively include factors such as initial investment, wind power price, battery life, etc., to optimize CCS capacity and to ultimately improve the operating indices of the joint system.

17 WIND ENERGY↗

DOE OSTI2016

Numerical simulations of depressurization-induced gas production from gas hydrates using 3-D heterogeneous models of L-Pad, Prudhoe Bay Unit, North Slope Alaska

Gas production potential is estimated using a three-dimensional reservoir model based on gas hydrate deposits located in the Prudhoe Bay region of the Alaska's North Slope. The model incorporates two hydrate-bearing sand units using detailed reservoir geological and structural information obtained from past and recent drilling programs. Geostatistical porosity models, conditioned to log data from 78 wells drilled in the vicinity of the Prudhoe Bay “L-Pad,” were developed, providing 3D heterogeneity in porosity, porosity-dependent hydrate saturation, and intrinsic permeability. The simulations utilize both vertical and inclined wellbores to induce depressurization of the reservoir at a constant bottom-hole pressure. The results show the superior performance of the inclined well design. Average gas production rates during the first five years were ~6.0 × 10 4 ST m 3 /day (~2.1 MMSCF/day) and ~2.7 × 10 4 ST m 3 /day (~1.0 MMSCF/day) for incline and vertical wells, respectively. After 30 years 5.3 × 10 8 - 5.7 × 10 8 ST m 3 (18.6–20.2 BSCF) and 6.2 × 10 8 - 6.4 × 10 8 ST m 3 (22.0–22.7 BSCF) gas were produced using the vertical and inclined well configurations, respectively. The analysis reveals that 2D reservoir models with homogeneous representations for porosity and hydrate saturation significantly underestimate production potential. The heterogeneity implemented in this work provides complex porous network and preferential pathways for mobile phase flow to a producing well. Consequently, no secondary hydrate formation around a wellbore is predicted in contrast to models that utilize uniform porosity and hydrate saturation distributions.

03 NATURAL GAS↗

DOE OSTI2018

Computation of the Seismic Stability of Earth Retaining Structures Final Report CRADA No. TSB-954-94

The purpose of this CRADA project was to evaluate the seismic stability of block retaining wall systems. Retaining wall systems are used extensively in private and commercial developments. This study was designed to develop and demonstrate a computer modeling technology to be used to predict the seismic stability of any block wall system design. The nonlinear finite element computer programs developed at LLNL and employed in the Computational Earthquake Initiative were utilized in this small business CRADA to analyze the seismic stability of the block retaining walls. The unique capability of the LLNL programs to rigorously model frictional contact in a dynamic analysis problem were used in a computer simulation of the dynamic interaction the block waiVsoil systems under seismic excitation.

58 GEOSCIENCES↗

DOE OSTI2023

UPC++ v1.0 Programmer’s Guide (Rev. 2023.9.0)

UPC++ is a C++ library that supports Partitioned Global Address Space (PGAS) programming. It is designed for writing efficient, scalable parallel programs on distributed-memory parallel computers. The key communication facilities in UPC++ are one-sided Remote Memory Access (RMA) and Remote Procedure Call (RPC). The UPC++ control model is single program, multiple-data (SPMD), with each separate constituent process having access to local memory as it would in C++. The PGAS memory model additionally provides one-sided RMA communication to a global address space, which is allocated in shared segments that are distributed over the processes. UPC++ also features Remote Procedure Call (RPC) communication, making it easy to move computation to operate on data that resides on remote processes. UPC++ was designed to support exascale high-performance computing, and the library interfaces and implementation are focused on maximizing scalability. In UPC++, all communication operations are syntactically explicit, which encourages programmers to consider the costs associated with communication and data movement. Moreover, all communication operations are asynchronous by default, encouraging programmers to seek opportunities for overlapping communication latencies with other useful work. UPC++ provides expressive and composable abstractions designed for efficiently managing aggressive use of asynchrony in programs. Together, these design principles are intended to enable programmers to write applications using UPC++ that perform well even on hundreds of thousands of cores.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

UPC++ v1.0 Programmer’s Guide (Revision 2022.3.0)

UPC++ is a C++ library that supports Partitioned Global Address Space (PGAS) programming. It is designed for writing efficient, scalable parallel programs on distributed-memory parallel computers. The key communication facilities in UPC++ are one-sided Remote Memory Access (RMA) and Remote Procedure Call (RPC). The UPC++ control model is single program, multiple-data (SPMD), with each separate constituent process having access to local memory as it would in C++. The PGAS memory model additionally provides one-sided RMA communication to a global address space, which is allocated in shared segments that are distributed over the processes. UPC++ also features Remote Procedure Call (RPC) communication, making it easy to move computation to operate on data that resides on remote processes. UPC++ was designed to support exascale high-performance computing, and the library interfaces and implementation are focused on maximizing scalability. In UPC++, all communication operations are syntactically explicit, which encourages programmers to consider the costs associated with communication and data movement. Moreover, all communication operations are asynchronous by default, encouraging programmers to seek opportunities for overlapping communication latencies with other useful work. UPC++ provides expressive and composable abstractions designed for efficiently managing aggressive use of asynchrony in programs. Together, these design principles are intended to enable programmers to write applications using UPC++ that perform well even on hundreds of thousands of cores.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

UPC++ v1.0 Programmer’s Guide, Revision 2023.3.0

UPC++ is a C++ library that supports Partitioned Global Address Space (PGAS) programming. It is designed for writing efficient, scalable parallel programs on distributed-memory parallel computers. The key communication facilities in UPC++ are one-sided Remote Memory Access (RMA) and Remote Procedure Call (RPC). The UPC++ control model is single program, multiple-data (SPMD), with each separate constituent process having access to local memory as it would in C++. The PGAS memory model additionally provides one-sided RMA communication to a global address space, which is allocated in shared segments that are distributed over the processes. UPC++ also features Remote Procedure Call (RPC) communication, making it easy to move computation to operate on data that resides on remote processes. UPC++ was designed to support exascale high-performance computing, and the library interfaces and implementation are focused on maximizing scalability. In UPC++, all communication operations are syntactically explicit, which encourages programmers to consider the costs associated with communication and data movement. Moreover, all communication operations are asynchronous by default, encouraging programmers to seek opportunities for overlapping communication latencies with other useful work. UPC++ provides expressive and composable abstractions designed for efficiently managing aggressive use of asynchrony in programs. Together, these design principles are intended to enable programmers to write applications using UPC++ that perform well even on hundreds of thousands of cores.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

UPC++ v1.0 Programmer’s Guide, Revision 2022.9.0

UPC++ is a C++ library that supports Partitioned Global Address Space (PGAS) programming. It is designed for writing efficient, scalable parallel programs on distributed-memory parallel computers. The key communication facilities in UPC++ are one-sided Remote Memory Access (RMA) and Remote Procedure Call (RPC). The UPC++ control model is single program, multiple-data (SPMD), with each separate constituent process having access to local memory as it would in C++. The PGAS memory model additionally provides one-sided RMA communication to a global address space, which is allocated in shared segments that are distributed over the processes. UPC++ also features Remote Procedure Call (RPC) communication, making it easy to move computation to operate on data that resides on remote processes. UPC++ was designed to support exascale high-performance computing, and the library interfaces and implementation are focused on maximizing scalability. In UPC++, all communication operations are syntactically explicit, which encourages programmers to consider the costs associated with communication and data movement. Moreover, all communication operations are asynchronous by default, encouraging programmers to seek opportunities for overlapping communication latencies with other useful work. UPC++ provides expressive and composable abstractions designed for efficiently managing aggressive use of asynchrony in programs. Together, these design principles are intended to enable programmers to write applications using UPC++ that perform well even on hundreds of thousands of cores.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Performance portable ice-sheet modeling with MALI

High-resolution simulations of polar ice sheets play a crucial role in the ongoing effort to develop more accurate and reliable Earth system models for probabilistic sea-level projections. These simulations often require a massive amount of memory and computation from large supercomputing clusters to provide sufficient accuracy and resolution; therefore, it has become essential to ensure performance on these platforms. Many of today’s supercomputers contain a diverse set of computing architectures and require specific programming interfaces in order to obtain optimal efficiency. In an effort to avoid architecture-specific programming and maintain productivity across platforms, the ice-sheet modeling code known as MPAS-Albany Land Ice (MALI) uses high-level abstractions to integrate Trilinos libraries and the Kokkos programming model for performance portable code across a variety of different architectures. In this article, we analyze the performance portable features of MALI via a performance analysis on current CPU-based and GPU-based supercomputers. The analysis highlights not only the performance portable improvements made in finite element assembly and multigrid preconditioning within MALI with speedups between 1.26 and 1.82x across CPU and GPU architectures but also identifies the need to further improve performance in software coupling and preconditioning on GPUs. We perform a weak scalability study and show that simulations on GPU-based machines perform 1.24–1.92x faster when utilizing the GPUs. The best performance is found in finite element assembly, which achieved a speedup of up to 8.65x and a weak scaling efficiency of 82.6% with GPUs. We additionally describe an automated performance testing framework developed for this code base using a changepoint detection method. The framework is used to make actionable decisions about performance within MALI. We provide several concrete examples of scenarios in which the framework has identified performance regressions, improvements, and algorithm differences over the course of 2 years of development.

54 ENVIRONMENTAL SCIENCES↗