Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Heterogeneous computing”

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

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

At least 217 records · Page 12
All sourcesNASA NTRSDOE OSTI
DOE OSTI2011

Challenges in Scheduling Aggregation in CyberPhysical Information Processing Systems

Data aggregation (a.k.a reduce operations) is an important element in information processing systems, including MapReduce clusters and cyberphysical networks. Unlike simple sensor networks, all the data in information processing systems must be eventually aggregated. Our goal is to lower overall latency in these systems by intelligently scheduling aggregation on intermediate routing nodes. Unlike previous models, our model explicitly takes into account link latency and computa- tion time. Our model also considers heterogeneous computing capabilities. In order to understand the potential challenges associated with constructing a distributed scheduler that minimizes la- tency, we ve developed a simulation of our model and tested the results of randomly scheduling nodes. Although these experiments were designed to provide data for a null-model, preliminary results have yielded a few interesting observations. We show that in cases where the computation time is larger than transmission time, in-network aggregation can have a large effect (reducing latency by 50% or more), but that naive scheduling can have a detrimental effect. Specifically, we show that when the root node (a.k.a the basestation) is faster than the other nodes, the latency can increase with increased coverage, and that these effects vary with the number of nodes present.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2011

Middleware Case Study: MeDICi

In many application domains in science and engineering, data produced by sensors, instruments and networks is naturally processed by software applications structured as a pipeline . Pipelines comprise a sequence of software components that progressively process discrete units of data to produce a desired outcome. For example, in a Web crawler that is extracting semantics from text on Web sites, the first stage in the pipeline might be to remove all HTML tags to leave only the raw text of the document. The second step may parse the raw text to break it down into its constituent grammatical parts, such as nouns, verbs and so on. Subsequent steps may look for names of people or places, interesting events or times so documents can be sequenced on a time line. Each of these steps can be written as a specialized program that works in isolation with other steps in the pipeline. In many applications, simple linear software pipelines are sufficient. However, more complex applications require topologies that contain forks and joins, creating pipelines comprising branches where parallel execution is desirable. It is also increasingly common for pipelines to process very large files or high volume data streams which impose end-to-end performance constraints. Additionally, processes in a pipeline may have specific execution requirements and hence need to be distributed as services across a heterogeneous computing and data management infrastructure. From a software engineering perspective, these more complex pipelines become problematic to implement. While simple linear pipelines can be built using minimal infrastructure such as scripting languages, complex topologies and large, high volume data processing requires suitable abstractions, run-time infrastructures and development tools to construct pipelines with the desired qualities-of-service and flexibility to evolve to handle new requirements. The above summarizes the reasons we created the MeDICi Integration Framework (MIF) that is designed for creating high-performance, scalable and modifiable software pipelines. MIF exploits a low friction, robust, open source middleware platform and extends it with component and service-based programmatic interfaces that make implementing complex pipelines simple. The MIF run-time automatically handles queues between pipeline elements in order to handle request bursts, and automatically executes multiple instances of pipeline elements to increase pipeline throughput. Distributed pipeline elements are supported using a range of configurable communications protocols, and the MIF interfaces provide efficient mechanisms for moving data directly between two distributed pipeline elements.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2014

Density-fitted singles and doubles coupled cluster on graphics processing units

We adapt an algorithm for singles and doubles coupled cluster (CCSD) that uses density fitting (DF) or Cholesky decomposition (CD) in the construction and contraction of all electron repulsion integrals (ERI s) for use on heterogeneous compute nodes consisting of a multicore CPU and at least one graphics processing unit (GPU). The use of approximate 3-index ERI s ameliorates two of the major difficulties in designing scientific algorithms for GPU s: (i) the extremely limited global memory on the devices and (ii) the overhead associated with data motion across the PCI bus. For the benzene trimer described by an aug-cc-pVDZ basis set, the use of a single NVIDIA Tesla C2070 (Fermi) GPU accelerates a CD-CCSD computation by a factor of 2.1, relative to the multicore CPU-only algorithm that uses 6 highly efficient Intel core i7-3930K CPU cores. The use of two Fermis provides an acceleration of 2.89, which is comparable to that observed when using a single NVIDIA Kepler K20c GPU (2.73).

Sherrill, David↗
DOE OSTI2015

Evaluating Performance Portability of OpenACC

Accelerator-based heterogeneous computing is gaining momentum in High Performance Computing arena. However, the increased complexity of the accelerator architectures demands more generic, high-level programming models. OpenACC is one such attempt to tackle the problem. While the abstraction endowed by OpenACC offers productivity, it raises questions on its portability. This paper evaluates the performance portability obtained by OpenACC on twelve OpenACC programs on NVIDIA CUDA, AMD GCN, and Intel MIC architectures. We study the effects of various compiler optimizations and OpenACC program settings on these architectures to provide insights into the achieved performance portability.

Sabne, Amit J↗
DOE OSTI2014

OpenARC: Extensible OpenACC Compiler Framework for Directive-Based Accelerator Programming Study

Directive-based, accelerator programming models such as OpenACC have arisen as an alternative solution to program emerging Scalable Heterogeneous Computing (SHC) platforms. However, the increased complexity in the SHC systems incurs several challenges in terms of portability and productivity. This paper presents an open-sourced OpenACC compiler, called OpenARC, which serves as an extensible research framework to address those issues in the directive-based accelerator programming. This paper explains important design strategies and key compiler transformation techniques needed to implement the reference OpenACC compiler. Moreover, this paper demonstrates the efficacy of OpenARC as a research framework for directive-based programming study, by proposing and implementing OpenACC extensions in the OpenARC framework to 1) support hybrid programming of the unified memory and separate memory and 2) exploit architecture-specific features in an abstract manner. Porting thirteen standard OpenACC programs and three extended OpenACC programs to CUDA GPUs shows that OpenARC performs similarly to a commercial OpenACC compiler, while it serves as a high-level research framework.

Lee, Seyong↗
DOE OSTI2018

DRAGON: breaking GPU memory capacity limits with direct NVM access

Heterogeneous computing with accelerators is growing in importance in high performance computing (HPC). Recently, application datasets have expanded beyond the memory capacity of these accelerators, and often beyond the capacity of their hosts. Meanwhile, nonvolatile memory (NVM) storage has emerged as a pervasive component in HPC systems because NVM provides massive amounts of memory capacity at affordable cost. Currently, for accelerator applications to use NVM, they must manually orchestrate data movement across multiple memories and this approach only performs well for applications with simple access behaviors. To address this issue, we developed DRAGON, a solution that enables all classes of GP-GPU applications to transparently compute on terabyte datasets residing in NVM. DRAGON leverages the page-faulting mechanism on the recent NVIDIA GPUs by extending capabilities of CUDA Unified Memory (UM). Our experimental results show that DRAGON transparently expands memory capacity and obtain additional speedups via automated I/O and data transfer overlapping.

Markthub, Pak↗
DOE OSTI2020

CCAMP: An Integrated Translation and Optimization Framework for OpenACC and OpenMP

Heterogeneous computing and exploration into specialized accelerators are inevitable in current and future supercomputers. Although this diversity of devices is promising for performance, the array of architectures presents programming challenges. High-level programming strategies have emerged to face these challenges, such as the OpenMP offloading model and OpenACC. The varying levels of support for these standards, however, within vendor-specific and open-source tools, as well as the lack of performance portability across devices, have prevented the standards from achieving their goals. To address these shortcomings, we present CCAMP, an OpenMP and OpenACC interoperable framework. CCAMP provides two primary facilities: language translation between the two standards and device-specific directive optimization within each standard. We show that by using the CCAMP framework, programmers can easily transplant non-portable code into new ecosystems for new architectures. Additionally, by using CCAMP device-specific directive optimizations, users can achieve optimized performance across architectures using a single source code.

Lambert, Jacob↗
DOE OSTI2022

COMPOFF: A Compiler Cost model using Machine Learning to predict the Cost of OpenMP Offloading

The HPC industry is inexorably moving towards an era of extremely heterogeneous architectures, with more devices configured on any given HPC platform and potentially more kinds of devices, some of them highly specialized. Writing a separate code suitable for each target system for a given HPC application is not practical. The better solution is to use directive-based parallel programming models such as OpenMP. OpenMP provides a number of options for offloading a piece of code to devices like GPUs. To select the best option from such options during compilation, most modern compilers use analytical models to estimate the cost of executing the original code and the different offloading code variants. Building such an analytical model for compilers is a difficult task that necessitates a lot of effort on the part of a compiler engineer. Recently, machine learning techniques have been successfully applied to build cost models for a variety of compiler optimization problems. In this paper, we present COMPOFF, a cost model which uses the multi-layer perceptrons to statically estimates the Cost of OpenMP OFFloading. We used six different transformations on a parallel code of Wilson Dslash Operator to support GPU offloading, and we predicted their cost of execution on different GPUs using COMPOFF during compile time. Our results show that this model can predict offloading costs with a root mean squared error in prediction of less than 0.5 seconds. Our preliminary findings indicate that this work will make it much easier and faster for scientists and compiler developers to port legacy HPC applications that use OpenMP to new heterogeneous computing environment.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Resource-Adaptive Federated Text Generation with Differential Privacy

In cross-silo federated learning (FL), sensitive text datasets remain confined to local organizations due to privacy regulations, making repeated training for each downstream task both communication-intensive and privacy-demanding. A promising alternative is to generate differentially private (DP) synthetic datasets that approximate the global distribution and can be reused across tasks. However, pretrained large language models (LLMs) often fail under domain shift, and federated finetuning is hindered by computational heterogeneity: only resource-rich clients can update the model, while weaker clients are excluded, amplifying data skew and the adverse effects of DP noise. We propose a flexible participation framework that adapts to client capacities. Strong clients perform DP federated finetuning, while weak clients contribute through a lightweight DP voting mechanism that refines synthetic text. To ensure the synthetic data mirrors the global dataset, we apply control codes (e.g., labels, topics, metadata) that represent each client’s data proportions and constrain voting to semantically coherent subsets. This two-phase approach requires only a single round of communication for weak clients and integrates contributions from all participants. Experiments show that our framework improves distribution alignment and downstream robustness under DP and heterogeneity.

Wang, Jiayi [ORNL]↗
DOE OSTI2019

PCS: A Productive Computational Science Platform

As modern supercomputers continue to be increasingly heterogeneous with diverse computational accelerators, software becomes a critical aspect necessary to fully exploit these accelerators while not increasing the effort and time required to make variable scientific discovery in the face of complicated programming environments introduced by these accelerators. This paper presents PCS, a productive computational science platform for cluster-scale heterogeneous computing. The PCS platform by design follows a basic principle that, multiple programming models in a unified programming environment, is essential to facilitate large-scale, accelerator-aware, heterogeneous computing for next- generation scientific applications. We propose a new scheme — Edge Neural Network — for managing regularity in large dataflow graphs using higher-order bulk synchronous operators, providing a foundation for formalizing computational models for cluster-wide dataflow programming of heterogeneous clusters.

36 MATERIALS SCIENCE↗
DOE OSTI2019

PCS: A Productive Computational Science Platform

As modern supercomputers continue to be increasingly heterogeneous with diverse computational accelerators, software becomes a critical aspect necessary to fully exploit these accelerators while not increasing the effort and time required to make variable scientific discovery in the face of complicated programming environments introduced by these accelerators. This paper presents PCS, a productive computational science platform for cluster-scale heterogeneous computing. The PCS platform by design follows a basic principle that, multiple programming models in a unified programming environment, is essential to facilitate large-scale, accelerator-aware, heterogeneous computing for next-generation scientific applications. We propose a new scheme — Edge Neural Network — for managing regularity in large dataflow graphs using higher-order bulk synchronous operators, providing a foundation for formalizing computational models for cluster-wide dataflow programming of heterogeneous clusters.

Ojika, David↗
DOE OSTI2019

PCS: A Productive Computational Science Platform

As modern supercomputers continue to be increasingly heterogeneous with diverse computational accelerators, software becomes a critical aspect necessary to fully exploit these accelerators while not increasing the effort and time required to make variable scientific discovery in the face of complicated programming environments introduced by these accelerators. This paper presents PCS, a productive computational science platform for cluster-scale heterogeneous computing. The PCS platform by design follows a basic principle that, multiple programming models in a unified programming environment, is essential to facilitate large-scale, accelerator-aware, heterogeneous computing for next- generation scientific applications. We propose a new scheme - Edge Neural Network - for managing regularity in large dataflow graphs using higher-order bulk synchronous operators, providing a foundation for formalizing computational models for cluster-wide dataflow programming of heterogeneous clusters.

36 MATERIALS SCIENCE↗
DOE OSTI2013

Computational Mechanics for Heterogeneous Materials

The subject of this work is the development of models for the numerical simulation of matter, momentum, and energy balance in heterogeneous materials. These are materials that consist of multiple phases or species or that are structured on some (perhaps many) scale(s). By computational mechanics we mean to refer generally to the standard type of modeling that is done at the level of macroscopic balance laws (mass, momentum, energy). We will refer to the flow or flux of these quantities in a generalized sense as transport. At issue here are the forms of the governing equations in these complex materials which are potentially strongly inhomogeneous below some correlation length scale and are yet homogeneous on larger length scales. The question then becomes one of how to model this behavior and what are the proper multi-scale equations to capture the transport mechanisms across scales. To address this we look to the area of generalized stochastic process that underlie the transport processes in homogeneous materials. The archetypal example being the relationship between a random walk or Brownian motion stochastic processes and the associated Fokker-Planck or diffusion equation. Here we are interested in how this classical setting changes when inhomogeneities or correlations in structure are introduced into the problem. Aspects of non-classical behavior need to be addressed, such as non-Fickian behavior of the mean-squared-displacement (MSD) and non-Gaussian behavior of the underlying probability distribution of jumps. We present an experimental technique and apparatus built to investigate some of these issues. We also discuss diffusive processes in inhomogeneous systems, and the role of the chemical potential in diffusion of hard spheres is considered. Also, the relevance to liquid metal solutions is considered. Finally we present an example of how inhomogeneities in material microstructure introduce fluctuations at the meso-scale for a thermal conduction problem. These fluctuations due to random microstructures also provide a means of characterizing the aleatory uncertainty in material properties at the mesoscale.

36 MATERIALS SCIENCE↗
DOE OSTI2026

ZENN: A thermodynamics-inspired computational framework for heterogeneous data–driven modeling

Traditional entropy-based methods—such as cross-entropy loss in classification problems—have long been essential tools for representing the information uncertainty and physical disorder in data and for developing artificial intelligence algorithms. However, the rapid growth of data across various domains has introduced new challenges, particularly the integration of heterogeneous datasets with intrinsic disparities. To address this, we introduce a zentropy-enhanced neural network (ZENN), extending zentropy theory into the data science domain via intrinsic entropy, enabling more effective learning from heterogeneous data sources. ZENN simultaneously learns both energy and intrinsic entropy components, capturing the underlying structure of multisource data. To support this, we redesign the neural network architecture to better reflect the intrinsic properties and variability inherent in diverse datasets. We demonstrate the effectiveness of ZENN on classification tasks and energy landscape reconstructions, showing its superior generalization capabilities and robustness-particularly in predicting high-order derivatives. In image and text classification tasks, ZENN demonstrates superior generalization by introducing a learnable temperature variable that models latent multisource heterogeneity, allowing it to surpass state-of-the-art models on CIFAR-10/100, BBC News, and AG News. As a practical application in materials science, we employ ZENN to reconstruct the Helmholtz energy landscape of Fe3Pt using data generated from density functional theory and capture key material behaviors, including negative thermal expansion and the critical point in the temperature–pressure space. Overall, this work presents a zentropy-grounded framework for data-driven machine learning, positioning ZENN as a versatile and robust approach for scientific problems involving complex, heterogeneous datasets.

36 MATERIALS SCIENCE↗
DOE OSTI2020

XACC: a system-level software infrastructure for heterogeneous quantum–classical computing

Quantum programming techniques and software have advanced significantly over the past five years, with a majority focusing on high-level language frameworks targeting remote REST library APIs. As quantum computing architectures advance and become more widely available, lower-level, system software infrastructures will be needed to enable tighter, co-processor programming and access models. In this work, we present XACC, a system-level software infrastructure for quantum–classical computing that promotes a service-oriented architecture to expose interfaces for core quantum programming, compilation, and execution tasks. Additionally, we detail XACC's interfaces, their interactions, and its implementation as a hardware-agnostic framework for both near-term and future quantum–classical architectures. We provide concrete examples demonstrating the utility of this framework with paradigmatic tasks. Our approach lays the foundation for the development of compilers, associated runtimes, and low-level system tools tightly integrating quantum and classical workflows.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Computational Modeling of Heterogeneity of Stress, Charge, and Cyclic Damage in Composite Electrodes of Li-Ion Batteries

Charge heterogeneity is a prevalent feature in many electrochemical systems. In a commercial cathode of Li-ion batteries, the composite is hierarchically structured across multiple length scales including the sub-micron single-crystal primary-particle domains up to the macroscopic particle ensembles. The redox kinetics of charge transfer and mass transport strongly couples with mechanical stresses. This interplay catalyzes substantial heterogeneity in the charge (re)distribution, stresses, and mechanical damage in the composite electrode during charging and discharging. We assess the heterogeneous electrochemistry and mechanics in a LiNi x Mn y Co z O 2 (NMC) cathode using a fully coupled electro-chemo-mechanics model at the cell level. A microstructure-resolved model is constructed based on the synchrotron X-ray tomography data. We calculate the stress field in the composite and then quantitatively evaluate the kinetics of surface charge transfer and Li transport biased by mechanical stresses. We further model the cyclic behavior of the cell. The repetitive deformation of the active particles and the weakening of the interfacial strength cause gradual increase of the interfacial debonding. The mechanical damage impedes electron transfer, incurs more charge heterogeneity, and results in the capacity degradation in batteries over cycles.

25 ENERGY STORAGE↗