Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Performance modeling”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

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

At least 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Performance Modeling of Tandem Photovoltaics: A Yearlong Outdoor Degradation Analysis of a Ga⁢As//Si Minimodule

We present a performance modeling and degradation analysis framework for tandem photovoltaic modules, building upon established procedures for crystalline silicon devices and adapting them to account for the spectral sensitivity of multijunction technologies. The methodology employs filter criteria to select outdoor measurements close to standard test conditions (STC) under stable spectral and ambient conditions, followed by normalization of power production data with corrections for temperature, irradiance, and precipitable water vapor. We demonstrate this framework using a mechanically stacked four-terminal gallium arsenide (Ga⁢As) // silicon (Si) tandem solar minimodule deployed outdoors from October 2019 to January 2021 in Golden, Colorado, USA. We determined degradation rates of −4.1 ±0.2%/year for the Ga⁢As subcell and −2.5 ±0.9%/year for the Si subcell, with analysis of individual performance metrics indicating that packaging degradation, particularly delamination, was the dominant failure mode. Simulations using PVcircuit, an open-source equivalent-circuit solver, confirmed these findings. The presented methodology provides a reproducible foundation for performance modeling and degradation analysis of emerging tandem technologies.

14 SOLAR ENERGY↗
DOE OSTI2022

Evaluating the Impact of Proprietary Oil & Gas Data on Machine Learning Model Performance Using a Quasi-Experimental Analytical Approach

This study implements a data-intensive supervised ML approach through a quasi-experimental framework with the objective of quantifying the impact of oil and gas operator-specific proprietary data on ML-based predictive model performance relative to using oil and gas datasets that may be more commonly publicly available. The models are designed to jointly predict daily oil, gas, and water production for horizontal wells as a function of bottom-hole pressure drawdown, spatial placement across the study domain, and well completion attributes. Model performance is quantified on holdout test data to evaluate how each dataset affects resulting model variant performance.

02 PETROLEUM↗
DOE OSTI2024

Multiscale Modeling of Silicon Carbide Cladding for Nuclear Applications: Thermal Performance Modeling

The complex multiscale and anisotropic nature of silicon carbide (SiC) ceramic matrix composite (CMC) makes it difficult to accurately model its performance in nuclear applications. The existing models for nuclear grade composite SiC do not account for the microstructural features and how these features can affect the thermal and structural behavior of the cladding and its anisotropic properties. In addition to the microstructural features, the properties of individual constituents of the composites and fiber tow architecture determine the bulk properties. Models for determining the relationship between the individual constituents’ properties and the bulk properties of SiC composites for nuclear applications are absent, although empirical relationships exist in the literature. Here, a hierarchical multiscale modeling approach was presented to address this challenge. This modular approach addressed this difficulty by dividing the various aspects of the composite material into separate models at different length scales, with the evaluated property from the lower-length-scale model serving as an input to the higher-length-scale model. The multiscale model considered the properties of various individual constituents of the composite material (fiber, matrix, and interphase), the porosity in the matrix, the fiber volume fraction, the composite architecture, the tow thickness, etc. By considering inhomogeneous and anisotropic contributions intrinsically, our bottom-up multiscale modeling strategy is naturally physics-informed, bridging constitutive law from micromechanics to meso-mechanics and structural mechanics. The effects that these various physical attributes and thermo-physical properties have on the composite’s bulk thermal properties were easily evaluated and demonstrated through the various analyses presented herein. Since silicon carbide fiber-reinforced SiC CMCs are also promising thermal–structural materials with a broad range of high-end technology applications beyond nuclear applications, we envision that the multiscale modeling method we present here may prove helpful in future efforts to develop and construct reinforced CMCs and other advanced composite nuclear materials, such as MAX phase materials, that can service under harsh environments of ultrahigh temperatures, oxidation, corrosion, and/or irradiation.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2021

HyPerPy (Hydrogen Extraction and Parabolic Trough Plant Performance Models using Python) [SWR-21-53]

The HyPerPy package consists of three scripts: 1) Model for Hydrogen Tracking in Parabolic Trough Power Plants, 2) Model for Receiver Performance, and 3) Model for Hydrogen Extraction Process. The power plant model (1) tracks hydrogen generation and transport within the circulating heat transfer fluid (HTF) of the power plant. This script is a transient, initial value simulation, in which the hydrogen concentration in the circulating HTF is 0 moles per cubic meter everywhere at time 0 seconds. Hydrogen concentration is calculated at discrete locations within the circulating HTF with 4-second time resolution. During each time step, the change in hydrogen concentration due to hydrogen generation and permeation is calculated for each location according to the local HTF temperature, vessel or piping properties, hydrogen concentration and partial pressure. This model predicts hourly hydrogen concentrations for typical operating days in the spring, summer, fall, and winter seasons. Data is used to create hourly mappings of hydrogen concentrations for a typical operating year. The receiver model (2) uses a 1-hour time step to estimate getter loading and annulus pressure. For each time step, the model uses HTF and ambient temperature data to estimate absorber tube, bellows, and getter temperatures for the time step. In addition, the model uses hourly HTF hydrogen concentrations that are generated by the power plant model (1), and annulus hydrogen pressure from the previous time step. With these data, the model calculates the moles of hydrogen permeating across the absorber tube and bellows during the time step. The net change in moles hydrogen is added or subtracted from the getter loading for the previous time step, and the hydrogen pressure is re-calculated based on getter loading and temperature. The model uses this algorithm to simulate hydrogen permeation and loading 24 hours per day, 365 days per year using seasonal temperature data. The model repeats these calculations for 25 years to create a mapping of receiver getter loading and annulus hydrogen pressure for four seasons of each year. The model for hydrogen extraction (3) estimates hydrogen extraction rates for a specific separation module configuration. The rate depends primarily on membrane area, vacuum pump performance, headspace gas flowrate to the membrane, and headspace gas hydrogen partial pressure. This model has two versions. The steady-state version predicts hydrogen extraction rates when the module is operating in separation mode. The dynamic version predicts hydrogen transfer through the membrane when the module is operating in sensor mode. The steady-state version is used with the plant model (1) to predict hydrogen partial pressures in the power plant when the extraction process is operating.

Glatzmaier, Gregory↗
DOE OSTI2021

Platform Agnostic Streaming Data Application Performance Models

The mapping of computational needs onto execution resources is, by and large, a manual task, and users are frequently guided simply by intuition and past experiences. We present a queueing theory based performance model for streaming data applications that takes steps towards a better understanding of resource mapping decisions, thereby assisting application developers to make good mapping choices. The performance model (and associated cost model) are agnostic to the specific properties of the compute resource and application, simply characterizing them by their achievable data throughput. We illustrate the model with a pair of applications, one chosen from the field of computational biology and the second is a classic machine learning problem.

Faber, Clayton↗
DOE OSTI2024

PV Performance Modeling and Stakeholder Engagement (Final Technical Report)

This core capability project’s objective is to increase the value of photovoltaic (PV) performance models by improving their functionality, demonstrating, and quantifying their validity, and offering a wide range of stakeholder engagement opportunities. In FY22-24, we developed new and improved modeling algorithms and functions to represent PV performance more accurately in a variety of environments and conditions. The “Model parameter toolkit” was developed and includes functions to translate between different module temperature models, incidence angle modifier models, and single-diode models. A new modeling capability named “PV Atlas” was also developed leveraging Sandia’s High Performance Computing resources. This capability allows us to investigate several questions and provide climate-specific best practices and geographic data files; all these are hosted on an interactive website on Sandia’s GitHub and can be used for training, system optimization, or to provide best practices for uncertainty reduction. For model validation, we published high-quality PV performance, and weather data; these data are well documented, filtered, and processed for quality and include examples on how to run PV simulations. We also developed well documented, standardized methods for validating PV models and ran independent model validation and 2 blind modeling intercomparisons engaging with 49 organizations from 17 countries. We co-led and contributed to a growing, well documented and maintained suite of open-source functions for PV modeling (i.e., the pvlib-python) and we outreached to the PV modeling stakeholders via the PVPMC workshops and web resources. In addition, this project supported US representation and leadership for the International Energy Agency (IEA) PVPS Task 13; specifically, members of our team led and supported 3 subtasks on: 1) Best practices for the optimization of bifacial photovoltaic tracking, 2) Extreme weather events and their multiple impact on PV power plants: Risks, failure mechanisms and mitigation strategies, and 3) Best practice guidelines for the use of economic and technical Key Performance Indicators (KPIs). This project resulted in the publications of 14 peer reviewed journal papers, 37 conference presentations, 6 SAND reports, 5 public datasets and 6 new webpages on the PVPMC website. It supported the release of 13 pvlib-python versions where 28 enhancements were from this PV Performance Modeling project. We co-organized 5 PVPMC workshops in FY22-24 with the participation of 214 unique institutions and around 700 participants. The PVPMC website was redesigned, and its reliability was improved; it receives over 50,000 visitors/year from 202 unique countries.

14 SOLAR ENERGY↗
DOE OSTI2022

Phase-field simulations to inform nuclear fuel performance modeling

Software tools to simulate nuclear fuel performance at the engineering scale, such as Idaho National Laboratory (INL)’s BISON code, are increasingly relied upon in regulatory and economic decision-making. However, accurate results from these tools depends on the availability of materials parameters that are used as input. In recent years, atomistic and mesoscale simulation methods have emerged as a cost-effective, expedient means to obtain such input parameters. Phase-field simulations using INL’s Marmot application have been used to obtain microstructure-level parameters and to improve material models for fuel performance modeling using BISON. In this talk, recent examples of this process are given, including applications in UO2, U3Si2, and UZr fuel.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2026

Advancing SiC clad fuel performance model: bridging micro- and macro-scale models and experimental validation

This report presents a workflow for advancing fuel performance modeling of SiC composite cladding for light-water reactors by linking microscale, experimental data-informed finite element analysis with rod-scale fuel performance codes such as BISON. The workflow uses X-ray computed tomography (XCT) to capture the actual geometry and processing-induced defects of as-fabricated SiC composite tube specimens, particularly porosity and wall-thickness variations, and converts the segmented XCT volumes into image-based finite element meshes for high fidelity structural analysis.

Koyanagi, Takaaki [Oak Ridge National Laboratory (↗
DOE OSTI2022

PyPVRPM: Photovoltaic Reliability and Performance Model in Python

The ability to perform accurate techno-economic analysis of solar photovoltaic (PV) systems is essential for bankability and investment purposes. Most energy yield models assume an almost flawless operation (i.e., no failures); however, realistically, components fail and get repaired stochastically. This package, PyPVRPM, is a Python translation and improvement of the Language Kit (LK) based PhotoVoltaic Reliability Performance Model (PVRPM), which was first developed at Sandia National Laboratories in Goldsim software (Granata et al., 2011) (Miller et al., 2012). PyPVRPM allows the user to define a PV system at a specific location and incorporate failure, repair, and detection rates and distributions to calculate energy yield and other financial metrics such as the levelized cost of energy and net present value (Klise, Lavrova, et al., 2017). Our package is a simulation tool that uses NREL’s Python interface for System Advisor Model (SAM) (National Renewable Energy Laboratory, 2020b) (National Renewable Energy Laboratory, 2020a) to evaluate the performance of a PV plant throughout its lifetime by considering component reliability metrics. Besides the numerous benefits from migrating to Python (e.g., speed, libraries, batch analyses), it also expands on the failure and repair processes from the LK version by including the ability to vary monitoring strategies. These failures, repairs, and monitoring processes are based on user-defined distributions and values, enabling a more accurate and realistic representation of cost and availability throughout a PV system’s lifetime.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Evaluation of distributed process-based hydrologic model performance using only a priori information to define model inputs

Fully distributed, integrated surface–subsurface hydrological models (ISSHMs) have seen renewed interest due to availability of better software, high performance computing facilities, and high-resolution, spatially extensive data products. ISSHMs are valuable as tools for advancing system understanding as they can resolve multiple processes defined on the plot scale including three-dimensional interaction of surface water and groundwater. Here, we evaluated the performance of an ISSHM, the Advanced Terrestrial Simulator (ATS), on seven diverse catchments across the continental US using widely available data products to define model inputs without calibration. We compare the ATS-simulated streamflow and evapotranspiration with gauge observations and MODIS-derived evapotranspiration, respectively. Using the Kling-Gupta Efficiency (KGE) as metric, ATS with default data products performed reasonably well at 6 of 7 catchments for streamflow. However, in one of those 6 catchments ATS had poor performance on baseflow and ATS’s overall performance was thus judged to be inadequate despite the acceptable KGE. ATS performance for evapotranspiration was good in all 7 catchments using default data products. In the two catchments where ATS streamflow performance using default data products was not acceptable, the performance was significantly improved by using local information on subsurface properties below the soil. We also compare the model-simulated streamflow and evapotranspiration with the Sacramento soil moisture accounting (SAC-SMA) model, a semi-distributed model that was calibrated on a catchment-by-catchment basis. Uncalibrated ATS performance is comparable to the calibrated SAC-SMA model in terms of streamflow while ATS performance is similar to or better (much better in certain catchments) in reproducing MODIS-derived evapotranspiration. Reasonably good performance of ATS without catchment-specific calibration provides new confidence in the ISSHM class of models and community data products as tools for advancing understanding of watershed function in a changing environment.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

A Performance Model of In-Situ Techniques

The computational capacity of High-Performance Computing (HPC) systems increases continuously with the rapid development of central processing units (CPUs) and graphic processing units (GPUs), while the in-/output (IO) subsystem develops relatively slowly and storage capacity is also limited. Data-intensive applications, which are designed to leverage the high computational capacity of HPC resources, typically generate a considerable amount of data for post-processing visualizations and data analytics. The limited IO speed and storage space could lead to constraints in the actual performance of these applications and, therefore, scientific discovery. In-situ techniques, where data is visualized/analysed while still in memory rather than through disk, can contribute to alleviating these problems as they can reduce or even fully avoid data writing/reading through the IO subsystem to/from storage. However, the overall efficiency of insitu techniques crucially depends on the characteristics of both the in-situ tasks and the applications, and the resource distribution among them. Therefore, choosing the right in-situ approach (synchronous, asynchronous, or hybrid) and resource allocation is essential to minimize overhead and maximize the benefits of concurrent execution. In this paper, we present a performance model of in-situ techniques to find the most beneficial in-situ approach and the preferred resource configuration. We verify the high accuracy of our approach with over 6800 measurements and provide use cases with different applications.

Ju, Yi [Max Planck Computing and Data Facility, Ga↗
DOE OSTI2023

Influence of Correlations on the Thermal Performance Modeling of Parabolic Trough Collectors

The influence of correlations on the thermal performance modeling of parabolic trough collectors was analyzed in this work. A versatile model for a parabolic trough collector was developed that allows one- and two-dimensional analysis and enables the use of correlations to calculate thermophysical properties and convection heat transfer coefficients. The model also allows the use of constant values for properties and/or coefficients obtained from the evaluation correlations at a specific temperature. The effect of each correlation was evaluated independently, and the results were compared with a reference case that considered a two-dimensional approach and used all the correlations. For the analyzed cases, the correlation for the absorber emittance has the strongest impact on the collector efficiency, leading to a lower error when used. Based on the results, a one-dimensional model approach considering a correlation for the absorber emittance leads to efficiency errors below 3% for collector lengths of up to 243.6 m. Compared with the reference case, a one-dimensional approach using all correlations for a collector with a length of 500 m, and operating with an inlet temperature of 773 K, can result in errors around 9%. However, using constant values for properties and heat transfer coefficients could lead to errors of up to 50%. Multiple thermal models for parabolic trough collectors proposed in the literature rely on a one-dimensional approach, estimated values for the heat transfer coefficients, and constant thermophysical properties. The errors associated with those approaches are analyzed and quantified in this work as a function of the collector length and operation temperature.

absorber emittance↗
DOE OSTI2020

Generation IV Benchmarking of TRISO Fuel Performance Models Under Accident Conditions Final Report

The Generation IV International Forum (GIF) is a co-operative international endeavor of fourteen members organized to carry out the research and development needed to establish the feasibility and performance capabilities of the next generation nuclear energy systems. GIF selected six reactor technologies, amongst which is the Very High Temperature Reactor (VHTR) that is primarily dedicated to the cogeneration of electricity and hydrogen. The technical basis for VHTR is the tristructural isotropic (TRISO)-coated particle fuel, the graphite as the core structure, helium coolant, as well as the dedicated core layout and lower power density to removal decay heat in a natural way. At the heart of safety features of the VHTR concept lie the TRISO fuel particles that are designed to keep their structural integrity and retain fission products at temperatures up to 1600°C. As part as the design and future operation of VHTRs, a key aspect is the accurate prediction of fuel performance under irradiation and accident conditions. Modeling and simulation allow prediction of TRISO fuel behavior when subject to neutron flux and in high temperature accident scenarios. The refinement of the fuel performance models and codes is performed by comparison to in-pile and out-of-pile experimental data that reproduce the expected irradiation conditions in high temperature gas-cooled reactors (HTGRs). Historically, the International Atomic Energy Agency (IAEA) developed a benchmark dedicated to the validation of predictive methods for fuel and fission product behavior through the Coordinated Research Program CRP-2 (IAEA, 1997). CRP-2 was later updated to cover fuel fabrication, quality assurance, irradiation performance, safety testing, and spent fuel. The scope of the resulting CRP-6 benchmarks focused on HTGR fuel performance and fission product release (IAEA, 2012). Taking advantage of additional TRISO fuel fabrication, irradiation, and safety testing campaigns, GIF launched a Generation IV Benchmarking of TRISO Fuel Performance Models under Accident Conditions in late 2015. This GIF benchmark is a three-year program steered by Idaho National Laboratory (INL, USA). The other participants include the Japan Atomic Energy Agency (JAEA, Japan) and the Korea Atomic Energy Research Institute (KAERI, Korea). The objectives of the benchmark are to: follow on the IAEA CRP benchmarks, (2) model fission product release under accident conditions, (3) compare results obtained by the fuel performance modeling codes of the benchmark participants, and (4) compare these code predictions to experimental data. Safety tests chosen for modeling include the first and second experiments of the Advanced Gas Reactor program (AGR-1 and AGR-2) and the High Flux Reactor (HFR) EU1bis experiment. This report presents the results obtained by the three research institutions using their respective fuel performance modeling codes. Comparisons of the corresponding fission product release predictions are made with experimental data from AGR-1, AGR-2, and HFR-EU1bis. The benchmark results show good agreements between all participants but also show a general trend of over-prediction of the experimental release data, which is mainly attributed to the use of over-estimated diffusion coefficients.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2021

Initial demonstration of automated fuel performance modeling with 1977 EBR-II metallic fuel pins using BISON code with FIPD and IMIS databases

Using the BISON fuel performance code, simulations were conducted using an automated process to read initial and operating conditions from the Fuels Irradiation and Physics Database (FIPD) and Integral Fast Reactor materials information system (IMIS) database, which contains metallic fuel data from the Experimental Breeder Reactor-II (EBR-II). This work demonstrates use of an integrated framework to access the vast majority of EBR-II experimental fuel pin data to support rapid development of fuel performance models for next-generation metallic fuel systems. With this capability, validation for fuel qualification can be performed rapidly. Between IMIS and FIPD, there is enough information to conduct 1977 unique EBR-II metallic fuel pin histories from 24 different experiments, at varying levels of detail between the two databases. Each of these histories includes a high-resolution power history, flux history, coolant channel flow rates, and coolant channel temperatures. Fission gas release (FGR), cumulative damage fraction (CDF), fuel axial swelling, cladding profilometry, and burnup were all simulated in BISON. The results were compared to post-irradiation examination (PIE) results for the initial demonstration of automated BISON modeling. BISON simulations conducted with IMIS and FIPD were in rough agreement with PIE measurements and calculations. Cladding profilometry, FGR, and fuel axial swelling were found to be in rough agreement with PIE measurements, depending on the physics used within the BISON input files. Here, the mechanical contact solver chosen was found to significantly impact axial fuel swelling and cladding strain predictions. CDF values were assessed to see whether pin failure may have been predicted (CDF ≥ 1). This work suggests that continued development of an automated tool for BISON should focus on inclusion of the Fast Flux Test Facility (FFTF) experimental data for a larger database for metallic fuel, improved physical models to better capture fuel performance, such as fuel-cladding interactions, and a more detailed comparison with available PIE data to further the BISON model development.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2026

Fuel Performance Modeling Internship Final Presentation

This study examines the performance of U-Zr and U-Pd-Zr annular metallic fuel rodlets and details the current status of modeling efforts regarding U-Pu-Zr solid metallic fuel rodlets which were experimentally evaluated as part of the Advanced Fuels Campaign (AFC). The AFC mission is to develop novel fuel technologies and facilitate the implementation of those technologies by industry partners. A key objective is to improve steady-state and transient performance over current fuel types. The experiments of interest in this study included metallic fuel rodlets within HT-9 cladding which were placed in SS-316 capsules and inserted in the Advanced Test Reactor (ATR). Certain mechanical and thermal conditions cannot be directly evaluated through experiments and fuel performance modeling is used to shed light on this evolution over time. In this study, BISON Multiphysics simulations are leveraged to investigate the state of the fuel system throughout and after the experimental conditions.

11 - NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2022

Representations and strategies for transferable machine learning improve model performance in chemical discovery

Strategies for machine-learning (ML)-accelerated discovery that are general across material composition spaces are essential, but demonstrations of ML have been primarily limited to narrow composition variations. By addressing the scarcity of data in promising regions of chemical space for challenging targets such as open-shell transition-metal complexes, general representations and transferable ML models that leverage known relationships in existing data will accelerate discovery. Over a large set (~1000) of isovalent transition-metal complexes, we quantify evident relationships for different properties (i.e., spin-splitting and ligand dissociation) between rows of the Periodic Table (i.e., 3d/4d metals and 2p/3p ligands). We demonstrate an extension to the graph-based revised autocorrelation (RAC) representation (i.e., eRAC) that incorporates the group number alongside the nuclear charge heuristic that otherwise overestimates dissimilarity of isovalent complexes. To address the common challenge of discovery in a new space where data are limited, we introduce a transfer learning approach in which we seed models trained on a large amount of data from one row of the Periodic Table with a small number of data points from the additional row. We demonstrate the synergistic value of the eRACs alongside this transfer learning strategy to consistently improve model performance. Analysis of these models highlights how the approach succeeds by reordering the distances between complexes to be more consistent with the Periodic Table, a property we expect to be broadly useful for other material domains.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Modeling, Performance Assessment, and Nodal Data Analysis of TRISO-Fueled Systems with Shift

This technical report documents several enhancements to the Shift Monte Carlo (MC) code under the US Department of Energy (DOE) Nuclear Energy Advanced Modeling and Simulation (NEAMS) program in fiscal year (FY) 2022. Performance enhancements were added to Shift specifically for tristructural isotropic (TRISO)–fueled reactor systems and guided based on performance analysis in FY 2021. For the pebble performance model developed in previous studies, the runtime improved by ~ 91× compared to the original model and ~ 2× compared to the user-optimized model. Compared to Serpent, Shift is ~ 3× slower if Serpent delta-tracking is enabled but ~ 2× faster when delta-tracking is disabled. The multigroup cross section generation was improved through simplifying tally input definitions, porting several post-processing tally operations from Python scripts into the Shift code base, and accounting for production reactions in the scattering multiplicity. Progress was also made on two emerging capabilities: (1) the development of Titan (a Shift reactor physics user interface) and (2) initial investigation into path-length tallies for computing multigroup scattering matrices.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2022

Design and analysis of CXL performance models for tightly-coupled heterogeneous computing

Truly heterogeneous systems enable partitioned workloads to be mapped to the hardware that nets the best performance. However, current practice requires that inter-device communication between different vendors' hardware use host memory as an intermediary step. To date, there are no widely adopted solutions that allow accelerators to directly transfer data. A new cache-coherent protocol, CXL, aims to facilitate easier, fine-grained sharing between accelerators. In this work we analyze existing methods for designing heterogeneous applications that target GPUs and FPGAs working collaboratively, followed by an exploration to show the benefits of a CXL-enabled system. Specifically, we develop a test application that utilizes both an NVIDIA P100 GPU and a Xilinx U250 FPGA to show current communication limitations. From this application, we capture overall execution time and throughput measurements on the FPGA and GPU. We use these measurements as inputs to novel CXL performance models to show that using CXL caching instead of host memory results in a 1.31X speedup, while a more tightly-coupled pipelined implementation using CXL-enabled hardware would result in a speedup of 1.45X.

Cabrera, Anthony↗