Expression Isolation of Compiler-Induced Numerical Inconsistencies in Heterogeneous Code
Explore the source record for details and available documents.
Computer science↗
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.
Micro- and nano-X-ray computed-tomography: A step forward in the characterization of the pore network of a leached cement paste
Pore structure of leached cement pastes (w/c = 0.5) was studied for the first time from micro-scale down to the nano-scale by combining micro- and nano-X-ray computed tomography (micro- and nano-CT). This allowed assessing the 3D heterogeneity of the pore network along the cement profile (from the core to the altered layer) of almost the entire range of cement pore size, i.e. from capillary to gel pores. We successfully quantified an increase of porosity in the altered layer at both resolutions. Porosity is increasing from 1.8 to 6.1% and from 18 to 58% at the micro-(voxel = 1.81 μm) and nano-scale (voxel = 63.5 nm) respectively. The combination of both CT allowed to circumvent weaknesses inherent of both investigation scales. In addition the connectivity and the channel size of the pore network were also evaluated to obtain a complete 3D pore network characterization at both scales.
36 MATERIALS SCIENCE↗
Enabling machine learning-ready HPC ensembles with Merlin
With the growing complexity of computational and experimental facilities, many scientific researchers are turning to machine learning (ML) techniques to analyze large scale ensemble data. With complexities such as multi-component workflows, heterogeneous machine architectures, parallel file systems, and batch scheduling, care must be taken to facilitate this analysis in a high performance computing (HPC) environment. Here, we present Merlin, a workflow framework to enable large ML-friendly ensembles of scientific HPC simulations. By augmenting traditional HPC with distributed compute technologies, Merlin aims to lower the barrier for scientific subject matter experts to incorporate ML into their analysis. As a producer–consumer workflow model, Merlin enables multi-machine, cross-batch job, dynamically allocated yet persistent workflows capable of utilizing surge-compute resources. Key features of Merlin are a flexible HPC-centric interface, low per-task overhead, multi-tiered fault recovery, and a hierarchical sampling algorithm that allows for $\mathscr{O}$(N) task execution and $\mathscr{O}$(N ln N) task queuing to ensembles of millions of tasks. In addition to Merlin’s design, we test the algorithm’s performance in an HPC center and demonstrate the ability to enqueue 40 million simulations in 100 s, with a 30 millisecond per-task overhead that is independent of ensemble size. Finally, we describe some example applications that Merlin has enabled on leadership-class HPC resources, such as the ML-augmented optimization of nuclear fusion experiments and the calibration of infectious disease models to study the progression of and possible mitigation strategies for COVID-19.
97 MATHEMATICS AND COMPUTING↗
Reconciling Work Functions and Adsorption Enthalpies for Implicit Solvent Models: A Pt (111)/Water Interface Case Study
Implicit solvent models are a computationally efficient method of representing solid/liquid interfaces prevalent in electrocatalysis, energy storage, and materials science. However, electronic structure changes induced at the metallic surface by the dielectric continuum are not fully understood. To address this, we perform DFT calculations for the Pt(111)/water interface, in order to compare Poisson–Boltzmann continuum solvation methods with ab initio molecular dynamics (AIMD) simulations of explicit solvent. We show that the implicit solvent cavity can be parametrized in terms of the electric dipole moment change at the equilibrated explicit Pt/water interface to obtain the potential of zero charge (PZC). We also compare the accuracy of aqueous enthalpies of adsorption of phenol on Pt(111) using geometry and charge density based dielectric cavitation methods. The ability to parametrize the cavity according to individual atoms, as afforded in the geometry based approach, is key to obtaining accurate enthalpy changes of adsorption under aqueous conditions. Additionally, we show that the electronic structure changes induced by explicit solvent and our proposed implicit solvent parametrization scheme yield comparable density difference profiles and d-band projected density of states. We therefore demonstrate the capability of implicit solvent approaches to capture both the energetics of adsorption processes and the main electronic effects of aqueous solvent on the metallic surface. Therefore, this work provides a scheme for computationally efficient simulations of interfacial processes for applications in areas such as heterogeneous catalysis and electrochemistry.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Fine-Tuning the Activity of Metal–Organic Framework Supported Cobalt Catalysts for the Oxidative Dehydrogenation of Propane
Few-atom cobalt-oxide clusters, when dispersed on a Zr-based metal–organic framework (MOF) NU-1000, have previously been shown to be active for the oxidative dehydrogenation (ODH) of propane at low temperatures (< 230 °C), affording a selective and stable propene production catalyst. In our current work, a series of promoter ions with varying Lewis acidity, including Ni(II), Zn(II), Al(III), Ti(IV) and Mo(VI), are anchored as metal-oxide,hydroxide clusters to NU-1000 via SIM (solvothermal deposition within MOFs–specifically the nodes) followed by incorporation of Co(II) ions via vapor-phase AIM (atomic layer deposition (ALD) in MOFs). This process yields a series of NU-1000-supported bimetallic-oxo,hydroxo,aqua clusters. Using difference envelope density (DED) analyses, the spatial locations of the promoter ions and catalytic cobalt ions are determined. For all samples the SIM-anchored promoter ions are sited between pairs of Zr 6 nodes along the MOF c-axis (channel-aligned axis) whereas the location of the AIM-anchored cobalt ions varies depending on the identity of promoter metal ion. With Ni(II)-, Al(III)-, or Ti(IV)-containing clusters as promoters, the oxy-cobalt species are sited atop the promoter sites; with Mo(VI) they grow exclusively on the MOF nodes sites (hexa-Zr(IV)- oxo,hydroxo,aqua units); with Zn(II) they grow on both the node and promoter. The NU-1000- supported bimetallic-oxide clusters are active for propane ODH after thermal activation under O 2 to open a cobalt coordination site and to oxidize Co(II) to Co(III), as evidenced by operando Xray absorption spectroscopy at the Co K-edge. The cobalt component is exclusively responsible for the observed catalysis. In accord with the decreasing Lewis acidity of the promoter ion, catalytic activity increases in the order: Mo(VI)<Ti(IV)<Al(III)<Zn(II)<Ni(II). The finding is attributed to increasing ease of formation Co(III)-O• species and stabilization of a cobalt(III)- oxyl/propane transition state as the Lewis acidity of the promoter ions decreases. (The promoterfree catalyst is sited on the oxy-Zr(IV) node of the MOF and is inferior catalytically to the strongly promoted versions of the catalyst.) The relative Lewis acidities of the promoters are inferred from the initial pKa values of aqua complexes of the corresponding single metal ions. Departures from the (ODH-rate)/(promoter-pKa) correlation are paralleled by differences in cobalt catalyst siting or structure. The combined results point to an increasing ability to fine-tune, with close to atomic specificity, the structure-dependent activity of MOF-supported heterogeneous catalysts for gas phase chemical reactions. When coupled with mechanistic studies–computational or experimental–this ability may translate into informed prediction of improved catalysts for propane ODH and other chemical reactions.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Advanced Health Information Technology Analytic Framework and Application to Hazard Detection
Health Information Technology (HIT) aims to improve healthcare outcomes by organizing and analyzing various health-related data. With data accumulating at a staggering rate, the importance of real-time analytics has been increasing dramatically, shifting the focus of informatics from batch processing to streaming analytics. HIT is also facing unprecedented challenges in adapting to this new requirement and leveraging advanced IT technologies. This paper introduces a HIT data and compute platform that supports multi-granularity real-time analytics from heterogeneous data sources. The paper first identifies functional requirements and proposes a framework that satisfies the requirements using state-of-the-art big data technologies including Apache Kafka, Spark Structured Streaming Engine, and Delta Lake. To demonstrate its capability to support data analytics in multiple time granularities analytics, a statistical process control-based hazard detection algorithm has been implemented on top of the framework to detect unexpected hazards from order cancellation data of the Department of US Veterans Affairs (VA) in near real-time.
Kumar, Mohit↗
Exploring Tradeoffs in Federated Learning on Serverless Computing Architectures
Federated learning is driving the development of new techniques to efficiently and securely use data across multiple sites while using diverse resources. One of these techniques is the use of the serverless computing paradigm to abstract away resource specific configurations, allowing federated learning across heterogeneous environments. However, deploying federated learning across edge resources, the cloud, and traditional HPC sites will require specialized approaches in order to best account for the weaknesses and strengths of each resource. In this work, we explore the new tradeoffs presented by managing a federated learning task across heterogeneous resources and demonstrate these tradeoffs with experiments using a serverless federated learning framework.
Baughman, Matt↗
Material Needs and Measurement Challenges for Advanced Semiconductor Packaging: Understanding the Soft Side of Science
This Perspective builds upon insights from the National Institute of Standards and Technology (NIST)-organized workshop, “Materials and Metrology Needs for Advanced Semiconductor Packaging Strategies,” held at the 35th annual Electronics Packaging Symposium in Binghamton, NY, on September 5, 2024. It outlines critical challenges and opportunities related to polymer-based “soft” materials in advanced semiconductor packaging, with emphasis on polymer science, measurement science (metrology), and the strategic development of Research-Grade Test Materials (RGTMs). These efforts, led by the NIST CHIPS team, aim to advance the fundamental understanding of structure-property-processing relationships, promote standardized guidelines and innovative methods for material characterization, and accelerate the development, qualification, and adoption of next-generation packaging materials. The Perspective also distills key insights from the panel discussion with industry experts, emphasizing the need for close collaboration among materials scientists, process engineers, and metrology experts to enable a holistic strategy, further highlighting the importance of cross-sector partnerships among industry, academia, and government to address pressing challenges in packaging materials and processes.
97 MATHEMATICS AND COMPUTING↗
SU-E-QI-14: Quantitative Variogram Detection of Mild, Unilateral Disease in Elastase-Treated Rats
Purpose: Determining the presence of mild or early disease in the lungs can be challenging and subjective. We present a rapid and objective method for evaluating lung damage in a rat model of unilateral mild emphysema based on a new approach to heterogeneity assessment. We combined octree decomposition (used in three-dimensional (3D) computer graphics) with variograms (used in geostatistics to assess spatial relationships) to evaluate 3D computed tomography (CT) lung images for disease. Methods: Male, Sprague-Dawley rats (232 ± 7 g) were intratracheally dosed with 50 U/kg of elastase dissolved in 200 μL of saline to a single lobe (n=6) or with saline only (n=5). After four weeks, 3D micro-CT images were acquired at end expiration on mechanically ventilated rats using prospective gating. Images were masked, and lungs were decomposed to homogeneous blocks of 2×2×2, 4×4×4, and 8×8×8 voxels using octree decomposition. The spatial variance – the square of the difference of signal intensity – between all pairs of the 8×8×8 blocks was calculated. Variograms – graphs of distance vs. variance - were made, and data were fit to a power law and the exponent determined. The mean HU values, coefficient of variation (CoV), and the emphysema index (EI) were calculated and compared to the variograms. Results: The variogram analysis showed that significant differences between groups existed (p<0.01), whereas the mean HU (p=0.07), CoV (p=0.24), and EI (p=0.08) did not. Calculation time for the variogram for a typical 1000 block decomposition was ∼6 seconds, and octree decomposition took ∼2 minutes. Decomposing the images prior to variogram calculation resulted in a ∼700x decrease in time as compared to other published approaches. Conclusions: Our results suggest that the approach combining octree decomposition and variogram analysis may be a rapid, non-subjective, and sensitive imaging-based biomarker for quantitative characterization of lung disease.
60 APPLIED LIFE SCIENCES↗
Computational Approaches for the Dynamics of Structure Formation in Self-Assembling Polymeric Materials
Polymeric materials can assemble into a multitude of intricate nanoscale morphologies whose free energy differs by only a fraction of the thermal energy per molecule. Such small free-energy differences pose a challenge for modeling and simulation but also offer exciting opportunities to direct the assembly of such materials into morphologies that do not correspond to those of equilibrium bulk structures. Over the past decade, significant progress has been achieved in our ability to guide their self-assembly through the use of confinement, topographical or chemical patterns, and electric fields. In contrast, approaches to guide self-assembly by tailoring the dynamics of structure formation have received less attention. Finally, this review discusses opportunities and challenges of recently developed computational strategies to predict the dynamics of self-assembly of polymeric materials on the basis of the underlying free-energy landscape.
36 MATERIALS SCIENCE↗
Remembering the Work of Phillip L. Geissler: A Coda to His Scientific Trajectory
Phillip L. Geissler made important contributions to the statistical mechanics of biological polymers, heterogeneous materials, and chemical dynamics in aqueous environments. He devised analytical and computational methods that revealed the underlying organization of complex systems at the frontiers of biology, chemistry, and materials science. In this retrospective we celebrate his work at these frontiers.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Effects of Horizontal Resolution, Domain Size, Boundary Conditions, and Surface Heterogeneity on Coarse LES of a Convective Boundary Layer
Atmospheric properties in a convective boundary layer vary over a wide range of spatial scales and are commonly studied using large-eddy simulations (LES) in various configurations. We examine how the boundary layer depth and distribution of variability across scales are affected by LES grid spacing, domain size, inhomogeneity of surface properties, and external forcing. Two different setups of the Weather Research and Forecasting (WRF) model are analyzed. A semi-idealized configuration uses a periodic domain, flat surface, prescribed homogeneous surface heat fluxes, and horizontally uniform profiles of large-scale advective tendencies. A nested LES setup employs a larger domain and realistic initial and boundary conditions, including an interactive land surface model with representative topography and vegetation and soil types. Subdomains of identical size are analyzed for all simulations. Characteristic structure sizes are quantified using the variability scales L 50 and L 95 , defined such that features smaller than that contain 50% and 95% of the total variance, respectively. Progressive increase in L 50 from vertical velocity to temperature and moisture structures is systematically reproduced in all simulation configurations. This dependence of L 50 on the considered variable complicates the development of scale-aware parameterizations for models with grid spacing in the “terra incognita”. In simulations using a larger domain with heterogeneous surface properties, the development of internal mesoscale patterns significantly affects variance distributions inside analyzed subdomains. Sizes of boundary layer structures also strongly depend on the LES grid spacing and, in case of heterogeneous surface and topography, on location of the subdomain inside a larger computational domain.
54 ENVIRONMENTAL SCIENCES↗
ORCHA: A performance portability system for extreme heterogeneity
Heterogeneity is the prevalent trend in the rapidly evolving high-performance computing (HPC) landscape in both hardware and application software. The diversity in hardware platforms, currently comprising various accelerators and a future possibility of specializable chiplets, poses a significant challenge for scientific software developers aiming to harness optimal performance across different computing platforms while maintaining the quality of solutions when their applications are simultaneously growing more complex. Code synthesis and code generation can provide mechanisms to mitigate this challenge. We have developed a divide and conquer approach where different aspects of performance are handled by different stand-alone tools that are interfaced with the application through generated code. This portability system, ORCHA, enables users to configure and orchestrate their computations among available resources on a platform by specifying a high-level recipe, thereby permitting a many-to-many paradigm where each recipe results in a different variant of the application. The core design goal is to let users decide the application’s hardware mapping and orchestration by editing only the high-level recipe—without modifying the maintained source code or binding the application to a particular runtime system. Tools in ORCHA distribution are: CG-Kit for translating the recipe into an execution graph; Milhoja to execute the graph by orchestrating data and task movement among hardware resources; and Macroprocessor that enables users to define their own code-shorthand for higher composability and easier management of code variants. Additionally, the design of ORCHA permits tools to work in a plug-and-play mode where the application can build and run without CG-Kit and Milhoja, and either tool can be swapped out for other tools with similar capabilities by modifying the code generation portion of ORCHA. In this paper, we describe the design of ORCHA and the role that code-generation plays in isolating applications from tools. We demonstrate the breadth of configurations ORCHA enables with a case study in which an application configuration is realized on three distinct hardware mappings—a GPU-centric, a CPU/GPU balanced, and a CPU/GPU concurrent layouts by using different recipes.
Lee, Youngjun↗
Getting To Exascale: Applying Novel Parallel Programming Models To Lab Applications For The Next Generation Of Supercomputers
As supercomputing moves toward exascale, node architectures will change significantly. CPU core counts on nodes will increase by an order of magnitude or more. Heterogeneous architectures will become more commonplace, with GPUs or FPGAs providing additional computational power. Novel programming models may make better use of on-node parallelism in these new architectures than do current models. In this paper we examine several of these novel models – UPC, CUDA, and OpenCL –to determine their suitability to LLNL scientific application codes. Our study consisted of several phases: We conducted interviews with code teams and selected two codes to port; We learned how to program in the new models and ported the codes; We debugged and tuned the ported applications; We measured results, and documented our findings. We conclude that UPC is a challenge for porting code, Berkeley UPC is not very robust, and UPC is not suitable as a general alternative to OpenMP for a number of reasons. CUDA is well supported and robust but is a proprietary NVIDIA standard, while OpenCL is an open standard. Both are well suited to a specific set of application problems that can be run on GPUs, but some problems are not suited to GPUs. Further study of the landscape of novel models is recommended.
97 MATHEMATICS AND COMPUTING↗
Navier: Dataflow Architecture for Computation Chemistry
Navier’s objectives were two evaluate the use of emerging technologies, especially dataflow accelerators, for high-performance computing (HPC) applications, specifically in the domain of chemistry, and to develop a prototype software stack to support such applications. Navier builds on capabilities previously developed by synergistic projects, such as PNNL Data Model Convergence (DMC) LDRD Hardware Advanced Workflows (HAW) and DuOMO, as well as DOE ARIAA. Throughout its 18 months, the Navier team developed new capabilities and artifacts at all levels of the HW/SW stack, provided a seamless way to integrate novel computing architectures (Sambanova SN10 and Xilinx Versal AI) into an existing software stack, developed chemistry workflows, data analytics tools, and HPC molecular dynamics workflows that leverage the developed stack and PNNL institutional investments in emerging architectures. Navier also explored the use of active learning to accelerate a computational chemistry workflow for organic molecules on PNNL Junction cluster (in collaboration with AMD/Xilinx). Navier developed tools, methodologies, and studies for hardware software co-design and (sparse) dataflow accelerators that are composable and can be used together or separately. These methodologies are now used in other projects, such as DOE AMAIS and HPDA. This report describes Navier’s achievement, the developed tools and methodologies, and the research findings and conclusions.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Community Requirements Meta-Analysis: Characterizing Needs and Opportunities for HPDF
This High Performance Data Facility (HPDF) Project is creating a new scientific user facility to provide advanced infrastructure for data-intensive science, supporting the DOE’s Office of Science (SC) community. HPDF’s mission is to enable and accelerate scientific discovery by delivering state-of-the-art data management infrastructure, capabilities, and tools. This meta-analysis examines the needs of the breadth of the SC community, captured in publicly available community reports or mission documents. The meta-analysis identifies and provides initial characterization of fifteen core requirements for the HPDF Project team to consider during the conceptual design phase. The fifteen requirements illustrate how scientific work among SC communities requires modern, seamless user experiences across the ASCR Ecosystem to advance the use of large volumes of heterogeneous data. The scientific community requires support for the missing middle of compute between local and HPC to interactively and collaboratively use growing datasets. Data producers and end users will benefit from enhanced data catalogs and portals that improve data access through advanced search of well curated data. The fifteen requirements are examined here organized across five themes for discussion. Examples in each theme illustrate the array of scientific needs that convey the important role that the fully realized and operational High Performance Data Facility will be able to play as an integral part of the evolving ASCR Ecosystem. Our amalgamated data tables from ESnet reports demonstrate ranges to the volumes of data HPDF must be concerned with, but limitations are inherent to this meta-analysis (see Key Challenges & Limitations). Feedback and validation of these requirements along with additional details and emergent community requirements will be gathered through user research and design activities.
97 MATHEMATICS AND COMPUTING↗
Bridging the Homogeneous-Heterogeneous Divide: Modeling Spin for Reactivity in Single Atom Catalysis
Single atom catalysts (SACs) are emergent catalytic materials that have the promise of merging the scalability of heterogeneous catalysts with the high activity and atom economy of homogeneous catalysts. Computational, first-principles modeling can provide essential insight into SAC mechanism and active site configuration, where the sub-nm-scale environment can challenge even the highest-resolution experimental spectroscopic techniques. Nevertheless, the very properties that make SACs attractive in catalysis, such as localized d electrons of the isolated transition metal center, make them challenging to study with conventional computational modeling using density functional theory (DFT). For example, Fe/N-doped graphitic SACs have exhibited spin-state dependent reactivity that remains poorly understood. However, spin-state ordering in DFT is very sensitive to the nature of the functional approximation chosen. In this work, we develop accurate benchmarks from correlated wavefunction theory (WFT) for relevant octahedral complexes. We use those benchmarks to evaluate optimal DFT functional choice for predicting spin state ordering in small octahedral complexes as well as models of pyridinic and pyrrolic nitrogen environments expected in larger SACs. Using these guidelines, we determine Fe/N-doped graphene SAC model properties and reactivity as well as their sensitivities to DFT functional choice. Finally, we conclude with broad recommendations for computational modeling of open-shell transition metal single-atom catalysts.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
A Multi-Objective Approach for Optimizing Edge-Based Resource Allocation Using TOPSIS
Existing approaches for allocating resources on edge environments are inefficient and lack the support of heterogeneous edge devices, which in turn fail to optimize the dependency on cloud infrastructures or datacenters. To this extent, we propose in this paper OpERA, a multi-layered edge-based resource allocation optimization framework that supports heterogeneous and seamless execution of offloadable tasks across edge, fog, and cloud computing layers and architectures. By capturing offloadable task requirements, OpERA is capable of identifying suitable resources within nearby edge or fog layers, thus optimizing the execution process. Throughout the paper, we present results which show the effectiveness of our proposed optimization strategy in terms of reducing costs, minimizing energy consumption, and promoting other residual gains in terms of processing computations, network bandwidth, and task execution time. We also demonstrate that by optimizing resource allocation in computation offloading, it is then possible to increase the likelihood of successful task offloading, particularly for computationally intensive tasks that are becoming integral as part of many IoT applications such robotic surgery, autonomous driving, smart city monitoring device grids, and deep learning tasks. The evaluation of our OpERA optimization algorithm reveals that the TOPSIS MCDM technique effectively identifies optimal compute resources for processing offloadable tasks, with a 96% success rate. Moreover, the results from our experiments with a diverse range of use cases show that our OpERA optimization strategy can effectively reduce energy consumption by up to 88%, and operational costs by 76%, by identifying relevant compute resources.
97 MATHEMATICS AND COMPUTING↗