Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Computer architecture”

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 1,423 records · Page 79

Distributed Training and Optimization of Neural Networks

Deep learning models are yielding increasingly better performances thanks to multiple factors. To be successful, model may have large number of parameters or complex architectures and be trained on large dataset. This leads to large requirements on computing resource and turn around time, even more so when hyperparameter optimization is done (e.g. search over model architectures). While this is a challenge that goes beyond particle physics, we review the various ways to do the necessary computations in parallel, and put it in the context of high-energy physics.

Vlimant, Jean-Roch↗

ARES v1.x - Performance Portable Tool to Simulate Supernovae based on Parthenon Framework

Historically, codes for simulating supernovae (such as Arepo, FLASH or LEAFS) have been at the forefront of scientific high-performance computing to the immense computational resources required for full 3D simulations. However, given the shift towards heterogenous HPC architectures, many current-generation codes are at the risk of losing their competitiveness as they are only designed to run on homogeneous CPU-only systems. There exist several efforts to enable these codes for GPU’s, however, these efforts only consider specific architectures or vendors (e.g., implement only CUDA or HIP), limiting themselves to a small range of exascale computing systems. Frameworks such as Kokkos aim to provide a framework which is agnostic of the targeted architecture, enabling the development of performant and portable code. In the Ares code, we develop a performance portable tool to simulate supernovae based on the Parthenon Framework, which in turn uses Kokkos in the background. Here, the Parthenon Framework provides an interface to the underlying mesh-refinement routines, which form the backbone of our code. In addition, we incorporate the already existing Singularity-EOS toolkit to provide us with various equations of state, primarily the Helmholtz equation of state. We also include the JINA Reaclib as a basis for our nuclear network solver. Finally, we implement a gravity solver to complete the required physics. This setup will provide us with a minimal code base to simulate supernova in a similar style to the tried-and-tested Arepo code, but in a futureproof performance portable framework.

Lim, Hyun↗

Control and Information Systems for the National Ignition Facility

Orchestration of every National Ignition Facility (NIF) shot cycle is managed by the Integrated Computer Control System (ICCS), which uses a scalable software architecture running code on more than 1950 front-end processors, embedded controllers, and supervisory servers. The ICCS operates laser and industrial control hardware containing 66 000 control and monitor points to ensure that all of NIF’s laser beams arrive at the target within 30 ps of each other and are aligned to a pointing accuracy of less than 50 μm root-mean-square, while ensuring that a host of diagnostic instruments record data in a few billionths of a second. NIF’s automated control subsystems are built from a common object-oriented software framework that distributes the software across the computer network and achieves interoperation between different software languages and target architectures. A large suite of business and scientific software tools supports experimental planning, experimental setup, facility configuration, and post-shot analysis. Standard business services using open-source software, commercial workflow tools, and database and messaging technologies have been developed. An information technology infrastructure consisting of servers, network devices, and storage provides the foundation for these systems. Thus, this work is an overview of the control and information systems used to support a wide variety of experiments during the National Ignition Campaign.

42 ENGINEERING↗

Uncertain quantum computing futures and potential energy and physical resource impacts at scale

Considerable attention has recently focused on the vast energy and water demands of supercomputing, namely large-scale data centers that underpin artificial intelligence (AI), one of the great disruptors of contemporary society. Looking ahead some years from now, quantum computing is poised to disrupt established computing paradigms once again. Scientists and engineers are now working intensely to bring this century-old dream of physicists to fruition. Yet, as quantum computers begin to be integrated with classical supercomputing architectures, the implications for energy and physical resource use also need to be understood, especially how they compare to today’s AI data centers. These impacts have not yet been quantified by the research community – a notable gap in the literature, even if commercial-scale deployment of Quantum-Accelerated Computing Infrastructure (QuACI) is not expected for a few more years. This study is the first to conduct such an assessment. Using publicly available information from academic sources and private industry, we characterize multiple configurations of superconducting qubit-based, fault-tolerant quantum computers (FTQC) that could plausibly be deployed at scale in the 2030s and into the 2040s. By parameterizing these FTQC systems at a process level, we conduct a prospective scenario analysis to quantify their energy and physical resource needs. While these estimates are uncertain, given the current state of quantum technologies and their unknown future trajectories, important insights can already be drawn. One key finding is that while the electricity needs for a fleet of FTQCs are within the bounds of previous modeling studies that have explored high electricity demand futures, the needs for certain physical resources, namely water and helium-3, could pose bottlenecks to QuACI scale-up.

Computing↗

A performance model for GPUs with caches

To exploit the abundant computational power of the world's fastest supercomputers, an even workload distribution to the typically heterogeneous compute devices is necessary. While relatively accurate performance models exist for conventional CPUs, accurate performance estimation models for modern GPUs do not exist. This paper presents two accurate models for modern GPUs: a sampling-based linear model, and a model based on machine-learning (ML) techniques which improves the accuracy of the linear model and is applicable to modern GPUs with and without caches. We first construct the sampling-based linear model to predict the runtime of an arbitrary OpenCL kernel. Based on an analysis of NVIDIA GPUs' scheduling policies we determine the earliest sampling points that allow an accurate estimation. The linear model cannot capture well the significant effects that memory coalescing or caching as implemented in modern GPUs have on performance. We therefore propose a model based on ML techniques that takes several compiler-generated statistics about the kernel as well as the GPU's hardware performance counters as additional inputs to obtain a more accurate runtime performance estimation for modern GPUs. We demonstrate the effectiveness and broad applicability of the model by applying it to three different NVIDIA GPU architectures and one AMD GPU architecture. On an extensive set of OpenCL benchmarks, on average, the proposed model estimates the runtime performance with less than 7 percent error for a second-generation GTX 280 with no on-chip caches and less than 5 percent for the Fermi-based GTX 580 with hardware caches. On the Kepler-based GTX 680, the linear model has an error of less than 10 percent. On an AMD GPU architecture, Radeon HD 6970, the model estimates with 8 percent of error rates. As a result, the proposed technique outperforms existing models by a factor of 5 to 6 in terms of accuracy.

97 MATHEMATICS AND COMPUTING↗

VisIt: An End-User Tool for Visualizing and Analyzing Very Large Data

VisIt is a popular open source tool for visualizing and analyzing data. It owes its success to its foci of increasing data understanding, large data support, and providing a robust and usable product, as well as its underlying design that fits today's supercomputing landscape. In this short paper, we describe the VisIt project and its accomplishments. The VisIt project's three focal points - understanding data, large data, and delivering a product - together form a powerful environment for analyzing data from HPC simulations. It is used in varied ways: it enables visualization scientists, computational code developers, and the physicists that run these codes to perform a broad range of data understanding activities, including debugging, making movies, and exploring data. The user interface portion of its design provides a powerful paradigm for analyzing data while the data processing portion of its design is well suited for big data. This in turn has led to many successes: in scaling up to high levels of concurrency and large data sizes, in providing a 'home' for large data algorithms, in understanding how to best use supercomputers, and, most importantly, in helping users understand their data. Further, despite significant upcoming changes in supercomputing architecture, VisIt's future appears bright, as it enjoys vibrant user and developer communities.

99 GENERAL AND MISCELLANEOUS↗

LCLS Big Data Handling – How I Learned to Stop Worrying and Love the Data Deluge

Advanced data and computing systems are vital to Linac Coherent Light Source (LCLS) operations, data interpretation and overall scientific productivity. The transition to MHz-era operation marks a fundamental change in scale that requires new infrastructure and architectures to link LCLS to the required scale of computing needed for scientific interpretation. The LCLS-II Data System meets big data challenges by implementing configurable data reduction that can adapt to multiple science areas, real-time analysis frameworks to provide visualization and fast feedback, and the ability to transfer data to local and remote computational facilities for near real time analysis at the appropriate scale. Feature extracted information generated in the data analysis pipeline - at the edge, local compute, or remote High-Performance Computing (HPC) resources - can be used to steer experiments and inform user decisions during beam time. Artificial Intelligence and Machine Learning (AI/ML) techniques present new opportunities to rapidly analyse large datasets and direct experiments, but create new challenges in scaling, adaptability, complexity, and trustworthiness. We describe how the LCLS-II Data System architecture addresses its data-driven challenges in the areas of data acquisition, data processing, data management, and workflow orchestration to decrease the overall time-to-science and provide a vision for future developments.

artificial intelligence↗

Large Scale Computing and Storage Requirements for High Energy Physics

The National Energy Research Scientific Computing Center (NERSC) is the leading scientific computing facility for the Department of Energy's Office of Science, providing high-performance computing (HPC) resources to more than 3,000 researchers working on about 400 projects. NERSC provides large-scale computing resources and, crucially, the support and expertise needed for scientists to make effective use of them. In November 2009, NERSC, DOE's Office of Advanced Scientific Computing Research (ASCR), and DOE's Office of High Energy Physics (HEP) held a workshop to characterize the HPC resources needed at NERSC to support HEP research through the next three to five years. The effort is part of NERSC's legacy of anticipating users needs and deploying resources to meet those demands. The workshop revealed several key points, in addition to achieving its goal of collecting and characterizing computing requirements. The chief findings: (1) Science teams need access to a significant increase in computational resources to meet their research goals; (2) Research teams need to be able to read, write, transfer, store online, archive, analyze, and share huge volumes of data; (3) Science teams need guidance and support to implement their codes on future architectures; and (4) Projects need predictable, rapid turnaround of their computational jobs to meet mission-critical time constraints. This report expands upon these key points and includes others. It also presents a number of case studies as representative of the research conducted within HEP. Workshop participants were asked to codify their requirements in this case study format, summarizing their science goals, methods of solution, current and three-to-five year computing requirements, and software and support needs. Participants were also asked to describe their strategy for computing in the highly parallel, multi-core environment that is expected to dominate HPC architectures over the next few years. The report includes a section that describes efforts already underway or planned at NERSC that address requirements collected at the workshop. NERSC has many initiatives in progress that address key workshop findings and are aligned with NERSC's strategic plans.

43 PARTICLE ACCELERATORS↗

PAMOP Project: Petaflop Computations in Support of Experiments.

Our computation effort is primarily concentrated in support of current and future measurements being carried out at various synchrotron radiation facilities around the globe and for charge transfer experiments. In our work we solve the Schrödinger or Dirac equation for the appropriate collision problem using the R-matrix or R-matrix with pseudo-states approach from first principles. The time dependent lattice (TDL) method is also used in our work on charge exchange. A brief summary of the methodology and ongoing developments implemented in the R-matrix suite of Breit-Pauli and Dirac-Atomic R-matrix codes (DARC) is presented. We illustrate vividly the sophistication level of large scale petaflop computations necessary to model accurately the spectra currently being measured at various synchrotron radiation facilities. The new Cray XC40 architecture installed at HLRS is playing a vital role in our computational effort.

Advance Light Source↗

Approximate Weighted Matching On Emerging Manycore and Multithreaded Architectures

Graph matching is a prototypical combinatorial problem with many applications in computer science and scientific computing, but algorithms for computing optimal matchings are challenging to parallelize. Approximate matching algorithms provide an alternate route for parallelization, and in many contexts compute near-optimal matchings for large-scale graphs. We present sharedmemory parallel implementations for computing half-approximate weighted matching on state-of-the-art multicore (Intel Nehalem and AMD Magny-Cours), manycore (Nvidia Tesla and Nvidia Fermi) and massively multithreaded (Cray XMT) platforms. We provide two implementations: the first implementation uses shared work queues, and is suited to all these platforms; the second implementation is based on dataflow principles, and exploits the architectural features of the Cray XMT. Using a carefully chosen dataset that exhibits characteristics from a wide range of real-world applications, we show scalable performance across different platforms. In particular, for one instance of the input, an R-MAT graph (RMAT-G), we show speedups of: about 32 on 48 cores of an AMD Magny-Cours; 7 on 8 cores of Intel Nehalem; 3 on Nvidia Tesla and 10 on Nvidia Fermi relative to one core of Intel Nehalem; and 60 on 128 processors of Cray XMT. We demonstrate good weak and strong scaling for graphs with up to a billion edges using up to 12, 800 threads. Given the breadth of this work, we focus on simplicity and portability of software rather than excessive fine-tuning for each platform. To the best of our knowledge, this is the first such large-scale study of the half-approximate weighted matching problem on shared-memory platforms. Driven by the critical enabling role of combinatorial algorithms such as matching in scientific computing and the emergence of informatics applications, there is a growing demand to support irregular computations on current and future computing platforms. In this context, we evaluate the capability of emerging multithreaded platforms to tolerate latency induced by irregular memory access patterns, and to support fine-grained parallelism via light-weight synchronization mechanisms. By contrasting the architectural features of these platforms against the Cray XMT, which is specifically designed to support irregular memory-intensive applications, we delineate the impact of these choices on performance.

Halappanavar, Mahantesh↗

Embedded Fluidic Sensing and Control with Soft Open‐Cell Foams

Abstract The synthesis of soft matter intelligence with circuit‐driven logic has enabled a new class of robots that perform complex tasks or conform to specialized form factors in unique ways that cannot be realized through conventional designs. Translating this hybrid approach to fluidic systems, the present work addresses the need for sheet‐based circuit materials by leveraging the innate porosity of foam—a soft material—to develop pneumatic components that support digital logic, mixed‐signal control, and analog force sensing in wearables and soft robots. Analytical tools and experimental techniques developed in this work serve to elucidate compressible gas flow through porous sheets, and to inform the design of centimeter‐sized foam resistors with fluidic resistances on the order of 10 9 Pa s m −3 . When embedded inside soft robots and wearables, these resistors facilitate diverse functionalities spanning both sensing and control domains, including digital logic using textile logic gates, digital‐to‐analog signal conversion using ladder networks, and analog sensing of forces up to 40 N via compression‐induced changes in resistance. By combining features of both circuit‐based and materials‐based approaches, foam‐enabled fluidic circuits serve as a useful paradigm for future hybrid robotic architectures that fully embody the sensing and computing capabilities of soft fluidic materials.

Rajappan, Anoop↗

Experimental Characterization of OpenMP Offloading Memory Operations and Unified Shared Memory Support

The OpenMP specification recently introduced support for unified shared memory, allowing implementation to leverage underlying system software to provide a simpler GPU offloading model where explicit mapping of variables is optional. Support for this feature is becoming more available in different OpenMP implementations on several hardware platforms. A deeper understanding of the different implementation’s execution profile and performance is crucial for applications as they consider the performance portability implications of adopting a unified memory offloading programming style. This work introduces a benchmark tool to characterize unified memory support in several OepnMP compilers and runtimes, with emphasis on identifying discrepancies between different OpenMP implementations as to how they various memory allocation strategies interact with unified shared memory. The benchmark tool is used to characterize OpenMP compilers on three leading High Performance Computing platforms supporting different CPU and device architectures. The benchmark tool is used to assess the impact of enabling unified shared memory on the performance of memory-bound code, highlighting implementation differences that should be accounted for when applications consider performance portability across platforms and compilers.

Elwasif, Wael↗

Toward a smart data transfer node

Scientific computing systems are becoming significantly more complex, with distributed teams and complex workflows spanning resources from telescopes and light sources to fast networks and Internet of Things sensor systems. In such settings, no single, centralized administrative team and software stack can coordinate and manage all resources used by a single application. Indeed, we have reached a critical limit in manageability using current human-in-the-loop techniques. We therefore argue that resources must begin to respond automatically, adapting and tuning their behavior in response to observed properties of scientific workflows. Over time, machine learning methods can be used to identify effective strategies for autonomic, goal-driven management behaviors that can be applied end-to-end across the scientific computing landscape. Using the data transfer nodes that are widely deployed in modern research networks as an example, we explore the architecture, methods, and algorithms needed for a smart data transfer node to support future scientific computing systems that self-tune and self-manage.

97 MATHEMATICS AND COMPUTING↗

Stepwise Bay Annulation of Indigo for the Synthesis of Desymmetrized Electron Acceptors and Donor–Acceptor Constructs

In this work, a selective stepwise annulation of indigo has been demonstrated as a means of providing both monoannulated and differentially double-annulated indigo derivatives. Disparate substitution of the electron accepting bay-annulated indigo system allows for fine control over both the electronic properties as well as donor-acceptor structural architectures. Optical and electronic properties were characterized computationally as well as through UV-vis absorption spectroscopy and cyclic voltammetry. Finally, this straightforward method provides a modular approach for the design of indigo-based materials with tailored optoelectronic properties.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Fast neural Poincaré maps for toroidal magnetic fields

Poincaré maps for toroidal magnetic fields are routinely employed to study gross confinement properties in devices built to contain hot plasmas. In most practical applications, evaluating a Poincaré map requires numerical integration of a magnetic field line, a process that can be slow and that cannot be easily accelerated using parallel computations. We propose a novel neural network architecture, the HénonNet, and show that it is capable of accurately learning realistic Poincaré maps from observations of a conventional field-line-following algorithm. After training, such learned Poincaré maps evaluate much faster than the field-line integration method. Moreover, the HénonNet architecture exactly reproduces the primary physics constraint imposed on field-line Poincaré maps: flux preservation. Furthermore, this structure-preserving property is the consequence of each layer in a HénonNet being a symplectic map. We demonstrate empirically that a HénonNet can learn to mock the confinement properties of a large magnetic island by using coiled hyperbolic invariant manifolds to produce a sticky chaotic region at the desired island location. This suggests a novel approach to designing magnetic fields with good confinement properties that may be more flexible than ensuring confinement using KAM tori.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Majorana subsystem qubit codes that also correct odd-weight errors

Abstract A potential platform for topological quantum computation is the Majorana-based tetron architecture. Its building blocks are superconducting islands called tetrons, which host four Majorana zero modes. Existing error correcting codes can correct even-weight errors on tetrons. In a previous proposal by us, we had shown that incorporating tetrons in the stabilizer group allows us to correct a combination of odd-weight errors and even-weight errors on tetrons. In this work, we show that inclusion of tetrons in the gauge group lets us create subsystem codes from conventional Pauli stabilizer codes, which can correct both kinds of errors. Compared to the previous approach, the current approach lets us construct codes with fewer stabilizer generators. This leads to shorter fault-tolerant sequence length, and improves the fault-tolerant pseudothreshold by as much as 84%.

Physics↗

Community detection robustness of graph neural networks

Graph neural networks (GNNs) are increasingly widely used for community detection in attributed networks. They combine structural topology with node attributes through message passing and pooling. However, their robustness or lack thereof with respect to different perturbations and targeted attacks in conjunction with community detection tasks is not well understood. To shed light on latent mechanisms behind GNN sensitivity on community detection tasks, we conduct a systematic computational evaluation of six widely adopted GNN architectures graph convolutional network, graph attention network, graph sample and aggregate (GraphSAGE), differentiable pooling (DiffPool), minimum cut pooling (MinCUT), and deep modularity networks (DMoN). The analysis covers three perturbation categories: node attribute manipulations, edge topology distortions, and adversarial attacks. We use element-centric similarity as the evaluation metric on synthetic benchmarks and real-world citation networks. Our findings indicate that supervised GNNs tend to achieve higher baseline accuracy, while unsupervised methods, particularly DMoN, maintain stronger resilience under targeted and adversarial perturbations. Furthermore, robustness appears to be strongly influenced by community strength, with well-defined communities reducing performance loss. Across all models, node attribute perturbations associated with targeted edge deletions and shifts in attribute distributions tend to cause the largest degradation in community recovery. These findings highlight important trade-offs between accuracy and robustness in GNN-based community detection and offer insights into selecting architectures resilient to noise and adversarial attacks.

Goel, Jaidev [Virginia Polytechnic Inst. and State↗

Deffe: A Data-efficient Framework for Performance Characterization in Domain-Specific Computing

Predicting workload performance is a crucial task for many architecture and system research. In this paper we present De￿e, a framework to estimate the workload performances under varying architectural configurations. The infrastructure component of Deffe is based on scalable and easy-to-use open-source software components. By casting the performance modeling as transfer learning tasks, the modeling component of Deffe can leverage the learned knowledge on one workload, and “transfer” it to a new workload. Extensive experimental results show that the method can achieve superior testing accuracy with an effective reduction of 32-80x in terms of the amount of required training data.

Liu, Frank↗