Search NASA⌕ Search

SEARCH · Search NASA

Results for “implements”

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

Implementation of a 1550-nm laser system for beam characterization at the Argonne Wakefield Accelerator

Accurately recording an electron bunch’s longitudinal profile is an important diagnostic for wakefield accelerators employing shaped bunches to increase transformer ratios. Electro-optic sampling of terahertz radiation from the bunch is an attractive approach due to its non-destructive nature. In preparation for future characterization experiments, the Argonne Wakefield Accelerator test facility has recently installed a 1550 nm laser system, including the necessary support systems to synchronize with the photoinjector laser system at 81.25 MHz. We report here on the initial installation and synchronization demonstrations.

Ody, Alexander [Argonne, PHY]↗
DOE OSTI2024

Implementation and Synthesis of Math Library Functions

Achieving speed and accuracy for math library functions like exp, sin, and log is difficult. This is because low-level implementation languages like C do not help math library developers catch mathematical errors, build implementations incrementally, or separate high-level and low-level decision making. This ultimately puts development of such functions out of reach for all but the most experienced experts. To address this, we introduce MegaLibm, a domain-specific language for implementing, testing, and tuning math library implementations. MegaLibm is safe, modular, and tunable. Implementations in MegaLibm can automatically detect mathematical mistakes like sign flips via semantic wellformedness checks, and components like range reductions can be implemented in a modular, composable way, simplifying implementations. Once the high-level algorithm is done, tuning parameters like working precisions and evaluation schemes can be adjusted through orthogonal tuning parameters to achieve the desired speed and accuracy. MegaLibm also enables math library developers to work interactively, compiling, testing, and tuning their implementations and invoking tools like Sollya and type-directed synthesis to complete components and synthesize entire implementations. MegaLibm can express 8 state-of-the-art math library implementations with comparable speed and accuracy to the original C code, and can synthesize 5 variations and 3 from-scratch implementations with minimal guidance.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Comparing computational times for simulations when using PBPK model template and stand-alone implementations of PBPK models

Introduction We previously developed a PBPK model template that consists of a single model “superstructure” with equations and logic found in many physiologically based pharmacokinetic (PBPK) models. Using the template, one can implement PBPK models with different combinations of structures and features. Methods To identify factors that influence computational time required for PBPK model simulations, we conducted timing experiments using various implementations of PBPK models for dichloromethane and chloroform, including template and stand-alone implementations, and simulating four different exposure scenarios. For each experiment, we measured the required computational time and evaluated the impacts of including various model features (e.g., number of output variables calculated) and incorporating various design choices (e.g., different methods for estimating blood concentrations). Results We observed that model implementations that treat body weight and dependent quantities as constant (fixed) parameters can result in a 30% time savings compared with options that treat body weight and dependent quantities as time-varying. We also observed that decreasing the number of state variables by 36% in our PBPK model template led to a decrease of 20–35% in computational time. Other factors, such as the number of output variables, the method for implementing conditional statements, and the method for estimating blood concentrations, did not have large impacts on simulation time. In general, simulations with PBPK model template implementations of models required more time than simulations with stand-alone implementations, but the flexibility and (human) time savings in preparing and reviewing a model implemented using the PBPK model template may justify the increases in computational time requirements. Conclusion Our findings concerning how PBPK model design and implementation decisions impact computational speed can benefit anyone seeking to develop, improve, or apply a PBPK model, with or without the PBPK model template.

Bernstein, Amanda S.↗
DOE OSTI2022

m-CUBES An efficient and portable implementation of multi-dimensional integration for gpus

The task of multi-dimensional numerical integration is frequently encountered in physics and other scientific fields, e.g., in modeling the effects of systematic uncertainties in physical systems and in Bayesian parameter estimation. Multi-dimensional integration is often time-prohibitive on CPUs. Efficient implementation on many-core architectures is challenging as the workload across the integration space cannot be predicted a priori. We propose m-Cubes, a novel implementation of the well-known Vegas algorithm for execution on GPUs. Vegas transforms integration variables followed by calculation of a Monte Carlo integral estimate using adaptive partitioning of the resulting space. m-Cubes improves performance on GPUs by maintaining relatively uniform workload across the processors. As a result, our optimized Cuda implementation for Nvidia GPUs outperforms parallelization approaches proposed in past literature. We further demonstrate the efficiency of m-Cubes by evaluating a six-dimensional integral from a cosmology application, achieving significant speedup and greater precision than the CUBA library's CPU implementation of VEGAS. We also evaluate m-Cubes on a standard integrand test suite. m-Cubes outperforms the serial implementations of the Cuba and GSL libraries by orders of magnitude speedup while maintaining comparable accuracy. Our approach yields a speedup of at least 10 when compared against publicly available Monte Carlo based GPU implementations. In summary, m-Cubes can solve integrals that are prohibitively expensive using standard libraries and custom implementations. A modern C++ interface header-only implementation makes m-Cubes portable, allowing its utilization in complicated pipelines with easy to define stateful integrals. Compatibility with non-Nvidia GPUs is achieved with our initial implementation of m-Cubes using the Kokkos framework.

Sakiotis, Ioannis↗
DOE OSTI2022

Field demonstration and implementation analysis of model predictive control in an office HVAC system

Model Predictive Control (MPC) is a promising technique to address growing needs for heating, ventilation, and air-conditioning (HVAC) systems to operate more efficiently and with greater flexibility. However, due to a number of factors, including the required implementation expertise, lack of high quality data, and a risk-adverse industry, MPC has yet to gain widespread adoption. While many previous studies have shown the advantages of MPC, few analyzed the implementation effort and associated practical challenges. In addition, previous work has developed an open-source, Modelica-based tool-chain that automatically generates optimal control, parameter estimation, and state estimation problems aimed at facilitating MPC implementation. Therefore, this study demonstrates usage of this tool-chain to implement MPC in a real office building, discusses practical challenges of implementing MPC, and estimates the implementation effort associated with various tasks in order to inform the development of future workflows and serve as an initial benchmark for their impact on reducing implementation effort. This study finds that the implemented MPC saves approximately 40% of HVAC energy over the existing control during a two-month trial period and that tasks related to data collection and controller deployment activities can each require as much effort as model generation.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2020

Taking stock of national climate policies to evaluate implementation of the Paris Agreement

Many countries have implemented national climate policies to accomplish pledged Nationally Determined Contributions and to contribute to the temperature objectives of the Paris Agreement on climate change. In 2023, the global stocktake will assess the combined effort of countries. Here, based on a public policy database and a multi-model scenario analysis, we show that implementation of current policies leaves a median emission gap of 22.4 to 28.2 GtCO 2 eq by 2030 with the optimal pathways to implement the well below 2 °C and 1.5 °C Paris goals. If Nationally Determined Contributions would be fully implemented, this gap would be reduced by a third. Interestingly, the countries evaluated were found to not achieve their pledged contributions with implemented policies (implementation gap), or to have an ambition gap with optimal pathways towards well below 2 °C. This shows that all countries would need to accelerate the implementation of policies for renewable technologies, while efficiency improvements are especially important in emerging countries and fossilfuel- dependent countries.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Implementation of ISO 15118-202 messages within Everest EV Charging Open Source Framework [SWR-25-56]

This software implements the messages defined in the ISO 15118-202 standard within the Everest EV Charging open source framework. The protocol and messages defined in the ISO 15118-202 standard enable the exchange of additional information which is not available for exchange within the currently deployed EV/EVSE communications protocols. This information includes co-identification parameters, error message exchange and more. This fork of the everest-core repository adds a prototype of the Extensible Supply Equipment Communication Controller (SECC) Discovery Protocol (ESDP) implemented based on a draft of the ISO 15118-202 standard. This is achieved through additions and modifications to the EvseV2G module. The implementation provides a demonstration of the ESDP messages, encoding and decoding but does not include a full integration within the Everest framework. Much of the information being sent over ESDP in this implementation is set statically for the sake of demonstrating the protocol itself. This fork of the ext-switchev-iso15118 repository adds a prototype of the Extensible Supply Equipment Communication Controller (SECC) Discovery Protocol (ESDP) implemented based on a draft of the ISO 15118-202 standard. The implementation provides a demonstration of the ESDP messages, encoding and decoding but does not include a full integration within the Everest framework. Much of the information being sent over ESDP in this implementation is set statically for the sake of demonstrating the protocol itself. This fork adds the ESDP features for only the EVCC controller because that is the only portion that is utilized in the everest Software-in-the-Loop.

Watt, Ed [National Renewable Energy Laboratory (NR↗
DOE OSTI2023

COVID-19 prevention at institutions of higher education, United States, 2020–2021: implementation of nonpharmaceutical interventions

Background, In early 2020, following the start of the coronavirus disease 2019 (COVID-19) pandemic, institutions of higher education (IHEs) across the United States rapidly pivoted to online learning to reduce the risk of on-campus virus transmission. We explored IHEs’ use of this and other nonpharmaceutical interventions (NPIs) during the subsequent pandemic-affected academic year 2020–2021. Methods, From December 2020 to June 2021, we collected publicly available data from official webpages of 847 IHEs, including all public (n = 547) and a stratified random sample of private four-year institutions (n = 300). Abstracted data included NPIs deployed during the academic year such as changes to the calendar, learning environment, housing, common areas, and dining; COVID-19 testing; and facemask protocols. We performed weighted analysis to assess congruence with the October 29, 2020, US Centers for Disease Control and Prevention (CDC) guidance for IHEs. For IHEs offering ≥50% of courses in person, we used weighted multivariable linear regression to explore the association between IHE characteristics and the summated number of implemented NPIs. Results, Overall, 20% of IHEs implemented all CDC-recommended NPIs. The most frequently utilized NPI was learning environment changes (91%), practiced as one or more of the following modalities: distance or hybrid learning opportunities (98%), 6-ft spacing (60%), and reduced class sizes (51%). Additionally, 88% of IHEs specified facemask protocols, 78% physically changed common areas, and 67% offered COVID-19 testing. Among the 33% of IHEs offering ≥50% of courses in person, having < 1000 students was associated with having implemented fewer NPIs than IHEs with ≥ 1000 students. Conclusions, Only 1 in 5 IHEs implemented all CDC recommendations, while a majority implemented a subset, most commonly changes to the classroom, facemask protocols, and COVID-19 testing. IHE enrollment size and location were associated with degree of NPI implementation. Additional research is needed to assess adherence to NPI implementation in IHE settings.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2023

Efficient GPU Implementation of Automatic Differentiation for Computational Fluid Dynamics

Many scientific and engineering applications require repeated calculations of derivatives of output functions with respect to input parameters. Automatic Differentiation (AD) is a method that automates derivative calculations and can significantly speed up code development. In Computational Fluid Dynamics (CFD), derivatives of flux functions with respect to state variables (Jacobian) are needed for efficient solutions of the nonlinear governing equations. AD of flux functions on graphics processing units (GPUs) is challenging as flux computations involve many intermediate variables that create high register pressure and require significant memory traffic because of the need to store the derivatives. This paper presents a forward-mode AD method based on multivariate dual numbers that addresses these challenges and simultaneously reduces the floating-point operation count. The dimension of the multivariate dual numbers is optimized for performance. The flux computations are restructured to minimize the number of temporary variables and reduce register pressure. For effective utilization of memory bandwidth, shared memory is used to store the local flux Jacobian. This AD implementation is compared with several other Jacobian implementations on an NVIDIA V100 GPU (V100). For three-dimensional perfect-gas compressible-flow equations implemented in a practical CFD code, the AD implementation of a flux Jacobian based on multivariate dual numbers of dimension 5 outperforms all other GPU AD implementations on V100. Its performance is comparable with the optimized hand-differentiated version. Finally, the implementation achieves 75% of the peak floating-point throughput and 61 % of the peak global device memory bandwidth usage.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2020

Implementation of early psychosis services in Latin America: A scoping review

Abstract Aim The evidence of the effectiveness and cost‐effectiveness of early intervention for psychosis (EIP) services has motivated their implementation worldwide. However, complex interventions of such EIP services require local adaptations to successfully match population needs and cultural differences. Latin America is a heterogenous region where EIP services are progressively being adopted. Our aim is to map such initiatives in the region with a focus on implementation outcomes. Methods A scoping review following the Preferred Reporting Items for Systematic review and Meta‐Analysis extension for Scoping Reviews guidelines was conducted. International and regional databases were searched for publications describing EIP programmes in the region. Besides mapping the services, we described implementation outcomes based on the Standards for Reporting Implementation Studies Checklist. Results Ten articles describing seven EIP initiatives from the region were found. Four countries were represented: Argentina, Brazil, Chile and Mexico. The implementation outcomes reporting was heterogenous, although it was possible to ascertain EIP services are feasible and adequate for the region's context. Also, there is some evidence of effectiveness in terms of reducing hospitalizations and improving symptoms. Information about fidelity measures was scarce and there was no information about costs or cost‐effectiveness. Conclusions Only a small proportion of Latin American countries have adopted EIP services. Although these programmes seem to be feasible and effective, data on other implementation outcomes, such as fidelity, cultural appropriateness, cost‐effectiveness and affordability are not available. This might in part explain why this effective approach has not been yet scaled‐up at nationwide levels.

Aceituno, David↗
DOE OSTI2025

Formulation and Experimental Validation of an Agricultural Implement-Only MPR System for Maximum Compatibility with Existing Agricultural Tractors

Tightening emissions regulations and rising fuel costs have driven a desire across many industries for more efficient actuation systems. This is particularly true of the agricultural sector. An extremely common arrangement in this sector is the tractor and implement pairing, in which actuators on an implement are powered by a hydraulic supply system on the towing tractor. This arrangement complicates the development of energy efficient hydraulic systems, as many new system designs require modification of both machines to reap efficiency benefits. Past work by the authors’ team has demonstrated great potentials for Multi-Pressure-Rail (MPR) technology involving both the tractor and implement subsystems. However, applicability of this MPR technology in a more realistic scenario where only one vehicle is equipped with such technology was not addressed. This work proposes an implementation of the MPR technology to an agricultural planter that allows significant savings, while only modifying the implement machine. This is done by manipulating the load sense network of a stock tractor to set system pressures to those required by the MPR system. This greatly reduces the barrier to implementation of MPR technology in agriculture. The work begins by outlining the reference machine for the system, then reviews the MPR system working principle. After this, the proposed expansion to the MPR concept is laid out and applied to the reference system. Finally, experimental validation is carried out, demonstrating up to a 35% reduction in system power consumption when paired with a state of the art, double-LS System tractor, and 15% with a single-LS System tractor.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

Using Generative AI to implement the discrepancy checker for a Nearly Autonomous Management and Control System for Advanced Reactors

Developments related to generative artificial intelligence (AI) have brought a major breakthrough in AI. These developments are rapidly accelerating developments in different science and engineering applications. Nearly Autonomous Management and Control (NAMAC) system provides recommendations to the operator for maintaining the safety and performance of the reactor. The discrepancy checker (DC) is an important component of the NAMAC) system, whose goal is to determine if the plant is moving towards the expected system state after the control actions are injected. In this work, we explore generative AI methods, particularly, a generative pretrained transformer (GPT) for implementing the DC function in NAMAC. The GPT-based DC aims to alert the operator in situations outside NAMAC’s scope and act as a chatbot the operator can use to retrieve relevant information. This study involves two versions of GPT developed by OpenAI: GPT-3.5 and GPT-4. These GPTs are trained on huge amounts of undisclosed general domain datasets. We explored two methods to adapt GPTs for DC implementation in NAMAC: fine-tuning and retrieval augmented generation. A small knowledge base (information file) that encompasses rules for DC implementation and some general information related to NAMAC has been created to support DC implementation using GPT. In this work, the GPT-based DC implementations have been tested for their reasoning abilities, comprehension, information retrieval, and extraction abilities. It should be noted that this paper only presents a preliminary study to test the feasibility of DC implementation using generative AI technology. Given the potential risks and severe consequences associated with nuclear reactor applications, combined with the black-box nature of AI, extensive offline and online testing and reliability analyses of GPT-based DCs are needed for further developing such capabilities.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2020

BSE versus StarTrack: Implementations of new wind, remnant-formation, and natal-kick schemes in NBODY7 and their astrophysical consequences

As a result of their formation via massive single and binary stellar evolution, the masses of stellar-remnant black holes (BH) are subjects of great interest in this era of gravitational-wave detection from binary black hole (BBH) and binary neutron star merger events. In this work, we present new developments in the stellar-remnant formation and related schemes of the current N-body evolution program NBODY7. We demonstrate that the newly implemented stellar-wind and remnant-formation schemes in the stellar-evolutionary sector or BSE of the NBODY7 code, such as the “rapid” and the “delayed” supernova (SN) schemes along with an implementation of pulsational-pair-instability and pair-instability supernova (PPSN/PSN), now produce neutron star (NS) and BH masses that agree nearly perfectly, over large ranges of zero-age-main-sequence (ZAMS) mass and metallicity, with those from the widely recognised StarTrack population-synthesis program. We also demonstrate the new, recipe-based implementations of various widely debated mechanisms of natal kicks on NSs and BHs, such as “convection-asymmetry-driven”, “collapse-asymmetry-driven”, and “neutrino-emission-driven” kicks, in addition to a fully consistent implementation of the standard, fallback-dependent, momentum-conserving natal kick. All the above newly implemented schemes are also shared with the standalone versions of SSE and BSE. Furthermore, all these demonstrations are performed with both the updated standalone BSE and the updated NBODY7/BSE. When convolved with stellar and primordial-binary populations as observed in young massive clusters, such remnant-formation and natal-kick mechanisms crucially determine the accumulated number, mass, and mass distribution of the BHs retained in young massive, open, and globular clusters (GCs); these BHs would eventually become available for long-term dynamical processing. Among other conclusions, we find that although the newer, delayed SN remnant formation model gives birth to the largest number (mass) of BHs, the older remnant-formation schemes cause the largest number (mass) of BHs to survive in clusters, when incorporating SN material fallback onto the BHs. The SN material fallback also causes the convection-asymmetry-driven SN kick to effectively retain similar numbers and masses of BHs in clusters as for the standard, momentum-conserving kick. The collapse-asymmetry-driven SN kick would cause nearly all BHs to be retained in clusters irrespective of their mass, remnant-formation model, and metallicity, whereas the inference of a large population of BHs in GCs would potentially rule out the neutrino-driven SN kick mechanism. Pre-SN mergers of massive primordial binaries would potentially cause BH masses to deviate from the theoretical, single-star ZAMS to mass-remnant mass relation unless a substantial of the total merging stellar mass of up to ≈40% is lost during a merger process. In particular, such mergers, at low metallicities, have the potential to produce low-spinning BHs within the PSN mass gap that can be retained in a stellar cluster and be available for subsequent dynamical interactions. As recent studies indicate, the new remnant-formation modelling reassures us that young massive and open clusters would potentially contribute to the dynamical BBH merger detection rate to a similar extent as their more massive GC counterparts.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

A Simple, Scalable Large Deformation Solid Mechanics Implementation in the MOOSE Framework

This article describes a large deformation solid mechanics solver implemented as part of the freely available and open source MOOSE finite element simulation framework. The article documents the choices made in developing the solid mechanics framework and describes novel formulations for the gradient operator and constitutive modeling framework made to simplify implementations of different coordinate systems, stabilized gradient operators, and different constitutive model inputs and outputs. In the process, the article describes a new formulation that casts objective integration of the Cauchy stress as a linear transformation of the small stress rate. Finally, the article presents key implementation details and examines the parallel efficiency of the solid mechanics solver implemented in MOOSE. The implementation retains a good weak scaling efficiency beyond 1,000 parallel processes. The article includes a discussion of the factors limiting the parallel efficiency of implicit, large deformation solid mechanics codes on current high-performance computers, with the main current limitation being the scalability of the algebraic multigrid methods used to solve the linearized equilibrium equations.

Applied computing → Computer-aided design↗
DOE OSTI2025

Implementation of Perturbation Theory and Sensitivity Capabilities in Griffin

Griffin is a Multiphysics Object-Oriented Simulation Environment (MOOSE) based reactor Multiphysics analysis application, jointly developed by Argonne and Idaho National Laboratories under the DOE-NE NEAMS program. This fiscal year, capabilities for reactivity and sensitivity evaluation using perturbation methods were implemented and verified. The First Order Perturbation Method (FOPT) was employed to compute reactivity worth resulting from small perturbations in input parameters, while the Generalized Perturbation Theory (GPT) was used to evaluate sensitivities of a range of response types, including reaction rate ratio, k-eigenvalue, neutron generation time, and effective delayed neutron fraction. These perturbation methods enable users to quantify how response quantities change due to a perturbation in a input parameter without explicitly performing an additional transport simulation for each perturbed state. In particular, the GPT formulation accounts for indirect effects arising from flux changes by solving generalized inhomogeneous equations, for which a Neumann series-based iterative solution method was developed and implemented in Griffin. The implemented reactivity and sensitivity evaluation capabilities were verified using two test problems: an infinite homogeneous system and a two-dimensional hexagonal core. The results showed excellent agreement with reference solutions obtained by a direct method based on finite difference approximation as well as GPT-based results from the PERSENT code, confirming the accuracy of both reactivity and sensitivity evaluations. Additionally, preliminary uncertainty quantification (UQ) results were obtained by combining the sensitivity values computed using GPT and external covariance data, demonstrating that the implemented sensitivity results can be reliably used for uncertainty calculations. To further demonstrate the generality and practical strength of the implementation, the sensitivity evaluation capability was successfully applied to the Empire microreactor with a geometrically complex design that poses significant modeling challenges. The results confirm that Griffin enables sensitivity evaluations even for irregular and highly heterogeneous reactor configurations, thereby establishing a foundation for UQ applications in advanced reactor designs and analyses.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2022

Efficient Implementation of Artificial Neural Networks for Sensor Data Analysis Based on a Genetic Algorithm

The reliability of many industrial processes depends on the sensor system. However, these sensors can be affected by noise, perturbations and failures. Hence, sensor monitoring and diagnosis are fundamental to guarantee the quality of an industrial process. Nowadays, artificial neural networks (ANN) are widely used in sensor signal processing and diagnosis. However, those ANNs usually require many artificial neurons, being difficult to implement in software and hardware due to their high computational costs. This paper presents an optimized implementation of artificial neurons in ANNs for sensor data analysis using a Genetic Algorithm (GA). The objective of GA is to find an adequate segmentation to reduce the activation function approximation error. One of the advantages of the proposed approach is that the cost function used in GA considers the effect of factors such as the ANN architecture or the number of bits used in arithmetic operations. The proposed ANN implementation technique aims to get the best possible approximation for a specific ANN architecture, making easier its implementation in software and hardware. Simulation and experimental results using FPGA (Field Programmable Gate Array) prove the advantages of the proposed approach for implementing sensor data analysis systems based on ANNs.

D estefani, André↗
DOE OSTI2023

Performance and Implementation Requirements for Residential EV Smart Charge Management Strategies

As the electrification of transportation expands, electric vehicle (EV) charging as residential loads will continue to grow. Residential EV charging has the potential to increase feeder peak loads and decrease voltage quality. As a result of this growing energy demand driven by EV, utilities may employ the use of smart charge management (SCM) controls to modify charging load profiles and mitigate these grid impacts. It is important that utilities understand both the potential benefits-as well as possible implementation challenges-before considering this technology as a solution to managing growing EV loads. In order for an SCM strategy to be an effective solution, the potential benefits must outweigh the implementation challenges. This study establishes and tests a novel framework to assess the implementation requirements of different SCM controls. It identifies a range of requirements specific to various SCM controls and implementation approaches to compare the relative challenges associated with the deployment of each. When paired with analysis on the effectiveness of the ability of each control to mitigate grid impacts from EV charging, this assessment is critical in comparing the value potential of different SCM controls.

ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2023

The Kokkos OpenMPTarget Backend: Implementation and Lessons Learned

As the supercomputing landscape diversifies, solutions such as Kokkos to write vendor agnostic applications and libraries have risen in popularity. Kokkos provides a programming model designed for performance portability, which allows developers to write a single source implementation that can run efficiently on various architectures. At its heart, Kokkos maps parallel algorithms to architecture and vendor specific backends written in lower level programming models such as CUDA and HIP. Another approach to writing vendor agnostic parallel code is using OpenMP’s directives based approach, which lets developers annotate code to express parallelism. It is implemented at the compiler level and is supported by all major high performance computing vendors, as well as the primary Open Source toolchains GNU and LLVM. Since its inception, Kokkos has used OpenMP to parallelize on CPU architectures. In this paper, we explore leveraging OpenMP for a GPU backend and discuss the challenges we encountered when mapping the Kokkos APIs and semantics to OpenMP target constructs. As an exemplar workload we chose a simple conjugate gradient solver for sparse matrices. We find that performance on NVIDIA and AMD GPUs varies widely based on details of the implementation strategy and the chosen compiler. Furthermore, the performance of the OpenMP implementations decreases with increasing complexity of the investigated algorithms.

Gayatri, Rahulkumar↗