SEARCH · Search NASA

Results for “Cluster 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.

SourceResearch connectionFrom yearThrough year

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

At least 685 records · Page 38

All sources NASA NTRS DOE OSTI

NASA NTRS1995

Computational strategies for three-dimensional flow simulations on distributed computer systems

This research effort is directed towards an examination of issues involved in porting large computational fluid dynamics codes in use within the industry to a distributed computing environment. This effort addresses strategies for implementing the distributed computing in a device independent fashion and load balancing. A flow solver called TEAM presently in use at Lockheed Aeronautical Systems Company was acquired to start this effort. The following tasks were completed: (1) The TEAM code was ported to a number of distributed computing platforms including a cluster of HP workstations located in the School of Aerospace Engineering at Georgia Tech; a cluster of DEC Alpha Workstations in the Graphics visualization lab located at Georgia Tech; a cluster of SGI workstations located at NASA Ames Research Center; and an IBM SP-2 system located at NASA ARC. (2) A number of communication strategies were implemented. Specifically, the manager-worker strategy and the worker-worker strategy were tested. (3) A variety of load balancing strategies were investigated. Specifically, the static load balancing, task queue balancing and the Crutchfield algorithm were coded and evaluated. (4) The classical explicit Runge-Kutta scheme in the TEAM solver was replaced with an LU implicit scheme. And (5) the implicit TEAM-PVM solver was extensively validated through studies of unsteady transonic flow over an F-5 wing, undergoing combined bending and torsional motion. These investigations are documented in extensive detail in the dissertation, 'Computational Strategies for Three-Dimensional Flow Simulations on Distributed Computing Systems', enclosed as an appendix.

Sankar, Lakshmi N.↗

NASA NTRS2012

The Gas Distribution in Galaxy Cluster Outer Regions

Aims. We present the analysis of a local (z = 0.04 - 0.2) sample of 31 galaxy clusters with the aim of measuring the density of the X-ray emitting gas in cluster outskirts. We compare our results with numerical simulations to set constraints on the azimuthal symmetry and gas clumping in the outer regions of galaxy clusters. Methods. We exploit the large field-of-view and low instrumental background of ROSAT/PSPC to trace the density of the intracluster gas out to the virial radius. We perform a stacking of the density profiles to detect a signal beyond r200 and measure the typical density and scatter in cluster outskirts. We also compute the azimuthal scatter of the profiles with respect to the mean value to look for deviations from spherical symmetry. Finally, we compare our average density and scatter profiles with the results of numerical simulations. Results. As opposed to some recent Suzaku results, and confirming previous evidence from ROSAT and Chandra, we observe a steepening of the density profiles beyond approximately r(sub 500). Comparing our density profiles with simulations, we find that non-radiative runs predict too steep density profiles, whereas runs including additional physics and/or treating gas clumping are in better agreement with the observed gas distribution. We report for the first time the high-confidence detection of a systematic difference between cool-core and non-cool core clusters beyond 0.3r(sub 200), which we explain by a different distribution of the gas in the two classes. Beyond r(sub 500), galaxy clusters deviate significantly from spherical symmetry, with only little differences between relaxed and disturbed systems. We find good agreement between the observed and predicted scatter profiles, but only when the 1% densest clumps are filtered out in the simulations. Conclusions. Comparing our results with numerical simulations, we find that non-radiative simulations fail to reproduce the gas distribution, even well outside cluster cores. Although their general behavior is in better agreement with the observations, simulations including cooling and star formation convert a large amount of gas into stars, which results in a low gas fraction with respect to the observations. Consequently, a detailed treatment of gas cooling, star formation, AGN feedback, and taking into account gas clumping is required to construct realistic models of cluster outer regions.

Eckert, D.↗

DOE OSTI2016

Measuring and Modeling Bipartite Graphs with Community Structure

Network science is a powerful tool for analyzing complex systems in fields ranging from sociology to engineering to biology. This paper is focused on generative models of bipartite graphs, also known as two-way graphs. We propose two generative models that can be easily tuned to reproduce the characteristics of real-world networks, not just qualitatively, but quantitatively. The measurements we consider are the degree distributions and the bipartite clustering coefficient, which we refer to as the metamorphosis coefficient. We define edge, node, and degreewise metamorphosis coefficients, enabling a more detailed understand of the bipartite community structure. Our proposed bipartite Chung-Lu model is able to reproduce real-world degree distributions, and our proposed bipartite “BTER” model reproduces both the degree distributions as well as the degreewise metamorphosis coefficients. We demonstrate the effectiveness of these models on several real-world data sets.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2018

Fundamental Vibrational Analyses of the HCN Monomer, Dimer and Associated Isotopologues

In this work we provide high level ab initio treatments of the structures, vibrational frequencies, and electronic energies of the HCN monomer and dimer systems along with several isotopologues. Additionally, the plethora of information related to this system within the literature is summarized and serves as a basis for comparison with the results of this paper. The geometry of the dimer and monomer are reported at the all electroncoupled-cluster singles, doubles, and perturbative triples level of theory [AE-CCSD(T)] with the correlation consistent quadruple-zeta quality basis sets with extra core functions (cc-pCVQZ) from Dunning. The theoretical geometries and electronic structures are further analyzed through the use of the Natural Bond Orbital (NBO) method and Natural Resonance Theory (NRT). At the AE-CCSD(T)/cc-pCVQZ level of theory, the full cubic with semi-diagonal quartic force field for nine dimer and four monomer isotopologues (the parent isotopologue along with 15N, 13C, and D derivatives) were obtained to treat the anharmonicity of the vibrations via second order vibrational perturbation theory (VPT2). Lastly, the enthalpy change associated with the formation of the dimer from two monomer units was determined using the focal point analysis. Computations including coupled-cluster through perturbative quadruples as well as basis sets up to six-zeta quality, including core functions (cc-pCVXZ, X=D,T,Q,5,6) were used to extrapolate to the AE-CCSDT(Q)/CBS energy associated with this hydrogen-bond forming process. After appending anharmonic zero-point vibrational, relativistic, and diagonal Born-Oppenheimer corrections, we report a value of -3.93 kcal mol -1 for the enthalpy of formation. To our knowledge, each set of results (geometries, vibrational frequencies, and energetics) reported in this study represents the highest-level and most reliable theoretical predictions reported for this system.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

NASA NTRS1996

Basic Requirements for Systems Software Research and Development

Our success over the past ten years evaluating and developing advanced computing technologies has been due to a simple research and development (R/D) model. Our model has three phases: (a) evaluating the state-of-the-art, (b) identifying problems and creating innovations, and (c) developing solutions, improving the state- of-the-art. This cycle has four basic requirements: a large production testbed with real users, a diverse collection of state-of-the-art hardware, facilities for evalua- tion of emerging technologies and development of innovations, and control over system management on these testbeds. Future research will be irrelevant and future products will not work if any of these requirements is eliminated. In order to retain our effectiveness, the numerical aerospace simulator (NAS) must replace out-of-date production testbeds in as timely a fashion as possible, and cannot afford to ignore innovative designs such as new distributed shared memory machines, clustered commodity-based computers, and multi-threaded architectures.

Kuszmaul, Chris↗

DOE OSTI2021

Global Optimization of Chemical Cluster Structures: Methods, Applications, and Challenges

Chemical clusters are relevant to many applications in catalysis, separations, materials, and energy sciences. Experimentally, the structure of clusters is difficult to determine, but it is very important in understanding their chemistry and properties. Computational methods can be used to examine cluster structure, however finding the most stable structure is not simple, particularly as the cluster size increases. Global optimization techniques have long been used to tackle the problem of the most stable structure, but such approaches would have to look for a global minimum, while sampling local minima over the whole potential energy surface as well. In this review, the state-of-the-art theory of global optimization theory is summarized. First, the definition, significance, relation to experiments, and a brief history of global optimization is presented. We then discuss, in more detail, three versatile global optimization methods: the basin hopping, the artificial bee colony algorithm, and the genetic algorithm. We close with some representative application examples of global optimization of clusters since 2016 and the challenges, open questions and opportunities in this field.

Global optimization, Chemical clusters, Artificial↗

DOE OSTI2020

Profiles of upcoming HPC Applications and their Impact on Reservation Strategies

With the expected convergence between HPC, BigData and AI, new applications with different profiles are coming to HPC infrastructures. Here, we aim at better understanding the features and needs of these applications in order to be able to run them efficiently on HPC platforms. The approach followed is bottom-up: we study thoroughly an emerging application from the neuroscience community (SLANT) to understand its behavior. Based on these observations, we derive a generic, yet simple, application model (namely, a linear sequence of stochastic jobs). We expect this model to be representative for a large set of upcoming applications that require the computational power of HPC clusters without fitting the typical behavior of large-scale traditional applications. In a second step, we show how one can manipulate this generic model in a scheduling framework. Specifically we consider the problem of making reservations (both time and memory) for an execution on an HPC platform. We derive solutions using the model of the first step of this work. We experimentally show the robustness of the model, even with very few data or with another application, to generate the model, and provide performance gains with regards to standard and more recent approaches used in the neuroscience community.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS2013

An Offload NIC for NASA, NLR, and Grid Computing

This work addresses distributed data management and access dynamically configurable high-speed access to data distributed and shared over wide-area high-speed network environments. An offload engine NIC (network interface card) is proposed that scales at nX10-Gbps increments through 100-Gbps full duplex. The Globus de facto standard was used in projects requiring secure, robust, high-speed bulk data transport. Novel extension mechanisms were derived that will combine these technologies for use by GridFTP, bandwidth management resources, and host CPU (central processing unit) acceleration. The result will be wire-rate encrypted Globus grid data transactions through offload for splintering, encryption, and compression. As the need for greater network bandwidth increases, there is an inherent need for faster CPUs. The best way to accelerate CPUs is through a network acceleration engine. Grid computing data transfers for the Globus tool set did not have wire-rate encryption or compression. Existing technology cannot keep pace with the greater bandwidths of backplane and network connections. Present offload engines with ports to Ethernet are 32 to 40 Gbps f-d at best. The best of ultra-high-speed offload engines use expensive ASICs (application specific integrated circuits) or NPUs (network processing units). The present state of the art also includes bonding and the use of multiple NICs that are also in the planning stages for future portability to ASICs and software to accommodate data rates at 100 Gbps. The remaining industry solutions are for carrier-grade equipment manufacturers, with costly line cards having multiples of 10-Gbps ports, or 100-Gbps ports such as CFP modules that interface to costly ASICs and related circuitry. All of the existing solutions vary in configuration based on requirements of the host, motherboard, or carriergrade equipment. The purpose of the innovation is to eliminate data bottlenecks within cluster, grid, and cloud computing systems, and to add several more capabilities while reducing space consumption and cost. Provisions were designed for interoperability with systems used in the NASA HEC (High-End Computing) program. The new acceleration engine consists of state-ofthe- art FPGA (field-programmable gate array) core IP, C, and Verilog code; novel communication protocol; and extensions to the Globus structure. The engine provides the functions of network acceleration, encryption, compression, packet-ordering, and security added to Globus grid or for cloud data transfer. This system is scalable in nX10-Gbps increments through 100-Gbps f-d. It can be interfaced to industry-standard system-side or network-side devices or core IP in increments of 10 GigE, scaling to provide IEEE 40/100 GigE compliance.

Awrach, James↗

DOE OSTI2023

Programmable photonic integrated meshes for modular generation of optical entanglement links

Abstract Large-scale generation of quantum entanglement between individually controllable qubits is at the core of quantum computing, communications, and sensing. Modular architectures of remotely-connected quantum technologies have been proposed for a variety of physical qubits, with demonstrations reported in atomic and all-photonic systems. However, an open challenge in these architectures lies in constructing high-speed and high-fidelity reconfigurable photonic networks for optically-heralded entanglement among target qubits. Here we introduce a programmable photonic integrated circuit (PIC), realized in a piezo-actuated silicon nitride (SiN)-in-oxide CMOS-compatible process, that implements an N × N Mach–Zehnder mesh (MZM) capable of high-speed execution of linear optical transformations. The visible-spectrum photonic integrated mesh is programmed to generate optical connectivity on up to N = 8 inputs for a range of optically-heralded entanglement protocols. In particular, we experimentally demonstrated optical connections between 16 independent pairwise mode couplings through the MZM, with optical transformation fidelities averaging 0.991 ± 0.0063. The PIC’s reconfigurable optical connectivity suffices for the production of 8-qubit resource states as building blocks of larger topological cluster states for quantum computing. Our programmable PIC platform enables the fast and scalable optical switching technology necessary for network-based quantum information processors.

47 OTHER INSTRUMENTATION↗

NASA NTRS2014

Vibrational Frequencies and Spectroscopic Constants for 1(sup 3)A' HNC and 1(sup 3)A' HOC+ from High-Accuracy Quartic Force Fields

The spectroscopic constants and vibrational frequencies for the 1(sup 3)A' states of HNC, DNC, HOC+, and DOC+ are computed and discussed in this work. The reliable CcCR quartic force field based on high-level coupled cluster ab initio quantum chemical computations is exclusively utilized to provide the anharmonic potential. Then, second order vibrational perturbation theory and vibrational configuration interaction methods are employed to treat the nuclear Schroedinger equation. Second-order perturbation theory is also employed to provide spectroscopic data for all molecules examined. The relationship between these molecules and the corresponding 1(sup 3)A' HCN and HCO+ isomers is further developed here. These data are applicable to laboratory studies involving formation of HNC and HOC+ as well as astronomical observations of chemically active astrophysical environments.

Astrochemistry↗

DOE OSTI2016

Accelerated Ion Irradiation for Emulation of In-reactor Radiation Damage

Reactor core materials in both fast reactors and LWRs granted life extension must withstand irradiation to high doses at high temperature. Ferritic-martensitic (F-M) alloys are attractive candidates for structural components and fuel cladding in fast and thermal reactors, and high chromium and high nickel-containing austenitic steels are potential replacement alloys for LWR core materials. To reach high doses, self-ion irradiation is conducted with simultaneous He injection into both F-M and austenitic alloys, accompanied by reactor irradiations conducted in the BOR-60 fast reactor to assess the capability of ion irradiation to emulate the evolution of microstructure and mechanical properties in reactor. Computational models for defect cluster evolution are being developed and benchmarked against experimental data to ultimately provide predictive capability for the response of both microstructure (loops, voids, precipitates, etc.), and mechanical properties (hardening, ductility, slip behavior) to irradiation. A research program is aimed at understanding how to emulate neutron irradiation using ion irradiation. Results to date have been focused on microstructure and properties of ion irradiated alloys T91, 800H and Fe21Ce32Ni. Data from single ion irradiations are thus far in good agreement with that expected based on previous ion and neutron irradiation data. Dual ion irradiations and neutron irradiated samples are being evaluated and these will be the focus of future work. (authors)

36 MATERIALS SCIENCE↗

DOE OSTI2021

Workflows Community Summit: Tightening the Integration between Computing Facilities and Scientific Workflows

Scientific workflows are used almost universally across science domains for solving complex and largescale computing and data analysis problems. The importance of workflows is highlighted by the fact that they have underpinned some of the most significant discoveries of the past decades. Many of these workflows have significant computational, storage, and communication demands, and thus must execute on a range of large-scale computer systems, from local clusters to public clouds and upcoming exascale HPC platforms. Managing these executions is often a significant undertaking, requiring a sophisticated and versatile software infrastructure. Historically, infrastructures for workflow execution consisted of complex, integrated systems, developed in-house by workflow practitioners with strong dependencies on a range of legacy technologies—even including sets of ad hoc scripts. Due to the increasing need to support workflows, dedicated workflow systems were developed to provide abstractions for creating, executing, and adapting workflows conveniently and efficiently while ensuring portability. While these efforts are all worthwhile individually, there are now hundreds of independent workflow systems. These workflow systems are created and used by thousands of researchers and developers, leading to a rapidly growing corpus of workflows research publications. The resulting workflow system technology landscape is fragmented, which may present significant barriers for future workflow users due to many seemingly comparable, yet usually mutually incompatible, systems that exist. In order to tackle some of the challenges described above, the DOE-funded ExaWorks and NSF-funded WorkflowsRI projects have organized in 2021 a series of events entitled the “Workflows Community Summit”. The third edition of the “Workflows Community Summit” explored workflows challenges and opportunities from the perspective of computing centers and facilities. This third summit builds on two prior summits (https://workflowsri.org/summits) that (i) established a high level vision for workflows research; and (ii) explored technical approaches for realizing that vision. The third summit brought together a small group of facilities representatives with the aim to understand how workflows are currently being used at each facility, how facilities would like to interact with workflow developers and users, how workflows fit with facility roadmaps, and what opportunities there are for tighter integration between facilities and workflows. This report documents and organizes the wealth of information provided by the participants before, during, and after the summit.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2015

Strong enhancement of surface diffusion by nonlinear surface acoustic waves

The phenomenon of acoustic activation of surface diffusion is investigated in a combined computational and experimental study. The ability of pulsed laser-generated surface acoustic waves (SAWs) to enhance the mobility of small atomic clusters is demonstrated by directly tracking, with fluorescence microscopy, individual Au 8 clusters moving on a (111) silicon substrate. In this work, a 19-fold increase in the effective diffusion coefficient is measured in room temperature experiments in the presence of SAWs generated by nanosecond pulse laser irradiation at a 100 Hz repetition rate. A strong enhancement of cluster mobility by SAWs is also observed in large-scale molecular dynamics simulations of surface diffusion of small atomic clusters. The analysis of the computational results demonstrates that the nonlinear sharpening of SAWs and the corresponding enrichment of the SAW spectra by high frequency harmonics which are capable of dynamic coupling to the cluster vibrations are responsible for the efficient acoustic activation of surface mobility in the simulations. The increase in the effective diffusion coefficient is proportional to the number of the SAW pulses passing through the diffusion region per unit time and a dramatic 4500-fold diffusion enhancement (corresponds to an equivalent temperature increase by ~10 3 K ) is predicted in the simulations for 15 GHz SAWs. The ability of SAWs to affect atomic-level surface processes has far-reaching implications for the design of new techniques where the acoustic energy serves as an effective substitution for thermal activation in applications where heating must be avoided or rapid switching of surface conditions is required.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

DOE OSTI2019

An Automated Thermochemistry Protocol Based on Explicitly Correlated Coupled-Cluster Theory: The Methyl and Ethyl Peroxy Families

An automated computational thermochemistry protocol based on explicitly correlated coupled-cluster theory was determined to produce highly accurate enthalpies of formation and atomization energies for small- to medium-sized molecular species (3–12 atoms). Each possible source of error was carefully examined, and the sizes of contributions to the total atomization enthalpies were used to generate uncertainty estimates. The protocol was first used to generate total atomization enthalpies for a family of four molecular species exhibiting a variety of charges, multiplicities, and electronic ground states. The new protocol was demonstrated to be in good agreement with the Active Thermochemical Tables database for the four species: the methyl peroxy radical, methoxyoxoniumylidene (methyl peroxy cation), methyl peroxy anion, and methyl hydroperoxide. Updating the Active Thermochemical Tables to include those results yielded significantly improved accuracy for the formation enthalpies of those species. The derived protocol was then used to predict formation enthalpies for the larger ethyl peroxy family of species.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2023

Determining Catalytically Relevant Surfaces through Coverage-Dependent Lattice Gas Models: Carbon Adsorption on Fe(100)

Here, we have quantified the C–C lateral interactions on Fe(100) using a density functional theory (DFT)-parameterized lattice gas cluster expansion (LG CE) model trained using 265 unique configurations spanning a C coverage from 0 to 1 monolayer (ML). Our LG CE model shows high predictive accuracy with a leave-multiple-out cross-validation score of 10.2 and 16.6 meV/site for systems with and without the top two layers of Fe atoms fixed, respectively. Electronic ground-state structures identified from the lattice gas model (including the structures at 0 and 1 monolayers) were further used to generate ab initio phase diagrams under a range of temperatures and pressures. At low temperatures (<400 K), we found that the 1.0 monolayer structure is dominant, whereas at higher temperatures (>500 K), the 0.88 ML structure is most likely to form on the Fe surface. Interestingly, our model identified a c (2 × 2) ordered structure at 1/2 ML, which correlates well with previous DFT studies for carbon adsorption on iron surfaces and matches with the experimentally observed low-energy electron diffraction structure. Overall, the DFT-parameterized energies for the C/Fe system including effects of coverage and configurational space can further help in developing multiscale models for various heterogeneous reactions involving C–C and C–Fe interactions.

08 HYDROGEN↗

DOE OSTI2022

Folding Coarse-Grained Oligomer Models with PyRosetta

Non-biological foldamers are a promising class of macromolecules that share similarities to classical biopolymers such as proteins and nucleic acids. Currently, designing novel foldamers is a non-trivial process, often involving many iterations of trial synthesis and characterization until folded structures are observed. In this work, we aim to tackle these foldamer design challenges using computational modeling techniques. We developed CG PyRosetta, an extension to the popular protein folding python package, PyRosetta, which introduces coarse-grained (CG) residues into PyRosetta, enabling the folding of toy CG foldamer models. Although these models are simplified, they can help explore overarching physical hypotheses about how oligomers can form. Through systematic variation of CG parameters in these models, we can investigate various folding hypotheses at the CG scale to inform the design process of new foldamer chemistries. In this study, we demonstrate CG PyRosetta’s ability to identify minimum energy structures with a diverse structural search over a range of simple models, as well as two hypothesis-driven parameter scans investigating the effects of side-chain size and internal backbone angle on secondary structures. We are able to identify several types of secondary structures from single- and double-helices to sheet-like and knot-like structures. Here, we show how side-chain size and backbone bond angle both play an important role in the structure and energetics of these toy models. Optimal side-chain sizes promote favorable packing of side chains, while specific backbone bond angles influence the specific helix type found in folded structures.

36 MATERIALS SCIENCE↗

NASA NTRS1993

A Grid-resolved Analysis of Base Flowfield for a Four-Engine Clustered Nozzle Configuration

The objective of this study is to propose a computational methodology that can effectively anchor the base flowfield of a four-engine clustered nozzle configuration. This computational methodology is based on a three-dimensional, viscous flow, pressure-based computational fluid dynamics (CFD) formulation. For efficient CFD calculation, a Prandtl-Meyer solution treatment is applied to the algebraic grid lines for initial plume expansion resolution. As the solution evolves, the computational grid is adapted to the pertinent flow gradients. The CFD model employs an upwind scheme in which second- and fourth-order central differencing schemes with artificial dissipation are used. The computed quantitative base flow properties such as the radial base pressure distributions, model centerline static pressure, Mach number and impact pressure variations, and base pressure characteristic curve agreed reasonably well with those of the measurement.

Wang, Ten-See↗

DOE OSTI2024

Acceleration of Graph Neural Network-Based Prediction Models in Chemistry via Co-Design Optimization on Intelligence Processing Units

Atomic structure prediction and associated property calculations are the bedrock of chemical physics. Since high-fidelity ab initio modeling techniques for computing the structure and properties can be prohibitively expensive, this motivates the development of machine-learning (ML) models that make these predictions more efficiently. Training graph neural networks over large atomistic databases introduces unique computational challenges such as the need to process millions of small graphs with variable size and support communication patterns that are distinct from learning over large graphs such as social networks. We demonstrate a novel hardware-software co-design approach to scale up the training of atomistic graph neural networks (GNN) for structure and property prediction. First, to eliminate redundant computation and memory associated with alternative padding techniques and to improve throughput via minimizing communication, we formulate the effective coalescing of the batches of variable-size atomistic graphs as the bin packing problem and introduce a hardware-agnostic algorithm to pack these batches. In addition, we propose hardware-specific optimizations including a planner and vectorization for the gather-scatter operations targeted for Graphcore’s Intelligence Processing Unit (IPU), as well as model-specific optimizations such as merged communication collectives and optimized softplus. Putting these all together, we demonstrate the effectiveness of the proposed co-design approach by providing an implementation of a well-established atomistic GNN on the Graphcore IPUs. We evaluate the training performance on multiple atomistic graph databases with varying degrees of graph counts, sizes and sparsity. Here, we demonstrate that such a co-design approach can reduce the training time of atomistic GNNs and can improve the performance by up to 1.5× compared to the baseline implementation of the model on the IPUs. Additionally, we compare our IPU implementation with a Nvidia GPU-based implementation and show that our atomistic GNN implementation on the IPUs can run 1.8× faster on average compared to the execution time on the GPUs.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗