Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “precondition”

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 649 records · Page 36
All sourcesNASA NTRSDOE OSTI
NASA NTRS2022

Understanding Structure in Line-Driven Stellar Winds Using Ultraviolet Spectropolarimetry in the Time Domain

The most massive stars are thought to lose a significant fraction of their mass in a steady wind during the main-sequence and blue supergiant phases. This in turn sets the stage for their further evolution and eventual supernova, and preconditions the surrounding medium for all following events, with consequences for ISM energization, chemical enrichment, and dust formation. Understanding these processes requires accurate observational constraints on the mass-loss rates of the most luminous stars, which can also be used to test theories of stellar wind driving. In the past, mass-loss rates have been characterized via collisional emission processes such as optical Hα and free-free radio emission, but these so-called “density squared” diagnostics require correction in the presence of widespread clumping. Recent observational and theoretical evidence points to the likelihood of a ubiquitously high level of such clumping in hot-star winds, but quantifying its effects requires a deeper understanding of the complex dynamics of radiatively driven winds and their stochastic instabilities. Furthermore, large-scale structures initiating in surface anisotropies and propagating throughout the wind can also affect wind driving and alter mass-loss diagnostics. Time series spectroscopy of high resonance-line opacity in the UV, capable of high resolution and high signal-to-noise, are required to better understand these complex dynamics, and more accurately determine mass-loss rates. The proposed Polstar mission (Scowen et al. 2022, this volume) provides the necessary resolution at the Sobolev (∼10 km s −1 ) or sound-speed (∼20 km s −1 ) scale, for over three dozen bright galactic massive stars with signal-to noise an order of magnitude above that of the celebrated MEGA campaign (Massa et al. 1995) of the International Ultraviolet Explorer (IUE), via continuous observations that track propagating structures through the winds in real time. Supporting geometric constraints are provided by the polarimetric capabilities present in all the datasets of such a mission.

Polstar – NASA MIDEX↗
NASA NTRS2022

The Polstar High Resolution Spectropolarimetry MIDEX Mission

The Polstar mission will provide a space-borne 60 cm spectropolarimeter operating at ultraviolet (UV) wavelengths, capturing all four Stokes parameters (intensity, two linear polarization components, and circular polarization). Polstar’s capabilities are designed to meet its goal of determining how circumstellar gas flows alter and inform massive star evolution, affect the stellar remnant population, and stir and enrich the interstellar medium (ISM). These will be achieved by investigating the dynamical geometries in the winds and disks of hot stars, the composition and magnetic alignment of interstellar dust, and the star-forming accretion disks of UV-bright stars at an important transition boundary. Together these areas map out a kind of two-way interface between massive stars and their effect on our galaxy, wherein the stellar winds enrich the ISM with metals and kinetic energy, preconditioning their environment and the stellar endpoints prior to undergoing supernova. The ISM dust in turn reveals the composition and magnetic environment leading to new star formation, and the accretion disks of Herbig Ae/Be stars reveal how the ISM gas returns to make new massive stars. Polstar will combine high-resolution spectroscopy in the time domain with high-precision UV polarimetry. Doppler-shifted UV resonance line opacity will provide information about circumstellar kinematics, while polarization gives complementary geometric information about unseen structures. The composition and magnetic alignment of the smallest interstellar dust grains provides a probe of the ISM utilizing radiative alignment theory (RAT). Polstar will operate in the far-UV (FUV) at 122–200 nm at high spectral resolution of around R ∼ 30k, and at FUV and near-UV (NUV) wavelengths of 122–320 nm at lower spectral resolutions of 0.1 - 1k. Detection of polarization levels as weak as 0.1% are expected, with a temporal cadence ranging from 5–10 minutes for most wind variability studies, to hours or days for sampling rotation, to days or weeks for sampling binary orbits, to months to a year for sampling substructure in the inner regions of protoplanetary disks. Sub-meter-class aperture is well suited to access this wide array of time domain science, made possible by restricting to a few hundred bright, massive stars, necessarily extincted by a small to moderate column of interstellar dust, informing both the attributes of the stars and the ISM through which they are seen. As such, the focus is on our own galaxy and its evolutionary drivers, but a few targets in the Magellanic clouds offer the potential to extend this understanding to low-metallicity environments.

Paul A Scowen↗
NASA NTRS2023

Recurrent 3he-Rich Solar Energetic Particle Injections Observed By Solar Orbiter at ∼0.5 Au

We report Solar Orbiter observations of six recurrent solar energetic particle injections in 2022 March 3−6 at ∼0.5 au. All but one were associated with jets emanating from a plage near a large sunspot in active region 12 957. We saw large jets in injections with high 3He and Fe enrichments and minor jets in injections with no or lower enrichments. Furthermore, the event with the highest enrichment showed a more compact configuration of the underlying photospheric magnetic field. The higher fluences as well as harder spectra were seen in the event with a simultaneous jet and wider eruption. However, in this case, the energy buildup time in the source might be required to produce such spectra. Extreme-ultraviolet images from Solar Orbiter revealed a number of intersecting loops at the base of jets not seen from 1 au that might be a precondition for the recurrent events.

Sun: particle emission↗
NASA NTRSDate not supplied

On the Prediction of Surface Melt Over the Greenland Ice Sheet at NWP and S2S Timescales

Greenland surface melt “events” – encompassing large swaths of the ice sheet over approximately 4-10 day periods – have occurred with increasing frequency in the 21st Century. Surface processes are significant to the ice sheet’s overall contribution to sea level. In addition, the sudden onset of widespread surface melt produces hazardous conditions for researchers and for communities adjacent to the ice sheet. The mechanisms producing melt events, their associated large-scale forcing, and timescales of interest are highly variable and poorly understood. Here we examine the predictability of melt events in a numerical weather prediction system (NWP) and a subseasonal-to-seasonal system (S2S). The NASA Global Modeling and Assimilation Office forward processing (GMAO-FP) is an NWP system that incorporates an evolving, global hybrid 4D ensemble–variational (4DEnVar) data assimilation system and a state-of-the-art model that is run twice-daily with forecasts out to five and ten days. The GMAO S2S system is a hindcast-anomaly, coupled-model prediction system initialized with an atmospheric analysis, and in situ and satellite ocean data assimilation. Both systems represent ice sheet surface hydrology and snowpack processes. We examine the performance of the systems over the period 2018-2023, a span that includes approximately nine significant events, against analyses and satellite-derived melt extent. In addition to extent, we examine related conditions and their role in the melt forecast, including the onset of both omega and Rex atmospheric blocking patterns, sea surface temperatures, sea ice extent, water vapor transport, precipitation, and surface preconditioning. We further examine the tendency for false-alarm events in each system and assess reasons for the forecast performance. North Atlantic atmospheric blocking patterns leading to surface melt are well forecast by the NWP system out to around eight days, while the S2S system generally underestimates the magnitude of such events at monthly lead times. These systems suggest that while short term forecasts of ice sheet melt and melt events are relatively robust, S2S predictions of surface conditions are highly dependent on the ability of a coupled system to generate the short-term atmospheric conditions associated with surface melt events.

Richard Cullather↗
NASA NTRS2024

Predictability of the Minimum Sea Ice Extent from Winter Fram Strait Ice Area Export: Model vs Observations

Observations show predictive skill of the minimum sea ice extent (Min SIE) from late winter anomalous off-shore ice drift along the Eurasian coastline – leading to local ice thickness anomalies at the onset of the melt season – a signal then amplified by the ice-albedo feedback. We assess whether the observed seasonal predictability of September sea ice extent (Sept SIE) from Fram Strait Ice Area Export (FSIAE, a proxy for Eurasian coastal divergence) is present in Global Climate Model (GCM) large ensembles, namely the CESM2-LE, GISS-E2.1-G, FLOR-LE, CNRM-CM6-1, and CanESM5. All models show distinct periods where winter FSIAE anomalies are negatively correlated with the May sea ice thickness (May SIT) anomalies along the Eurasian coastline, and the following Sept Arctic SIE, as in observations. Counter-intuitively, several models show occasional periods where winter FSIAE anomalies are positively correlated with the following Sept SIE anomalies when the mean ice thickness is large or late in the simulation when the sea ice is thin, and/or when internal variability increases. More importantly, periods with weak correlation between FSIAE and the Sept SIE dominate, suggesting that summer melt processes dominate over late winter preconditioning and May SIT anomalies. In general, we find that the coupling between the FSIAE and ice thickness anomalies along the Eurasian coastline is an ubiquitous feature of GCMs and that the relationship with the following Sept SIE is dependent on the mean Arctic sea ice thickness.

off-shore ice drift↗
DOE OSTI2021

Reliable Power Rating of Perovskite PV Modules: Preprint

As the perovskite technology is ramping up into commercialization, reliable and accurate power rating of large-size perovskite modules becomes a prominent aspect for its future deployment in the PV market. It is known that the performance calibration of perovskite PV devices is very challenging due to its complex dynamic response during a conventional current-voltage (IV) measurement. PV researchers have previously proposed several steady-state performance calibration methods to reliably extract PV efficiencies, but mostly focus on small area research-type cells. In this paper, we emphasize the importance of reliable performance calibration on large-size perovskite modules. Extending the NREL Cell and Module Performance (CMP) group’s steady-state performance calibration protocol (i.e., Asymptotic PMAX Scan) for perovskite cells to modules, we justify the necessity of reporting steady-state efficiencies for perovskite cells and discuss the challenges of applying this protocol to modules. We also present our protocol for Maximum Power Point Tracking (MPPT), which is a technique often used for performance calibration of perovskite cells and modules, and show a comparison between MPPT and Asymptotic PMAX. Using MPPT we demonstrate the interplay between metastability and degradation in perovskite modules, and emphasize the necessity to develop preconditioning protocols for stabilizing these devices. Our aim is to promote development of consensus protocols for performance calibration of perovskite modules, and to advance their credible power ratings, which will be beneficial to the growth of perovskite technology in the PV market.

41 EE - Solar Energy Technologies Office (EE-4S)↗
DOE OSTI2022

Two-Stage Gauss-Seidel Preconditioners and Smoothers for Krylov Solvers on a GPU Cluster: Preprint

Gauss-Seidel (GS) relaxation is often employed as a preconditioner for a Krylov solver or as a smoother for Algebraic Multigrid (AMG). However, the requisite sparse triangular solve is difficult to parallelize on many-core architectures such as graphics processing units (GPUs). In the present study, the performance of the sequential GS relaxation based on a triangular solve is compared with two-stage variants, replacing the direct triangular solve with a fixed number of inner Jacobi-Richardson (JR) iterations. When a small number of inner iterations is sufficient to maintain the Krylov convergence rate, the two-stage GS (GS2) often outperforms the sequential algorithm on many-core architectures. The GS2 algorithm is also compared with JR. When they perform the same number of ops for SpMV (e.g. three JR sweeps compared to two GS sweeps with one inner JR sweep), the GS2 iterations, and the Krylov solver preconditioned with GS2, may converge faster than the JR iterations. Moreover, for some problems (e.g. elasticity), it was found that JR may diverge with a damping factor of one, whereas two-stage GS may improve the convergence with more inner iterations. Finally, to study the performance of the two-stage smoother and preconditioner for a practical problem, these were applied to incompressible uid ow simulations on GPUs.

algebraic multigrid↗
DOE OSTI2022

Improvements to the Griffin Transport Solvers

Griffin is a Multiphysics Object-Oriented Simulation Environment (MOOSE) based reactor multiphysics analysis application jointly developed by Idaho National Laboratory and Argonne National Laboratory. The code includes a variety of steady-state solvers for fixed-source, k-eigenvalue, adjoint, and subcritical multiplication, as well as transient solvers for point-kinetics, improved quasi-static, and spatial dynamics. This document summarizes the transport solver development efforts pursued during Fiscal Year 2022. We added the multiphysics transient capability for the coarse-mesh finite difference accelerated Richardson iteration for discontinuous finite element method discrete ordinates (DFEM-SN) scheme to support high-order heterogeneous transport simulations. HFEM (hybrid finite element method) - PN (spherical harmonics expansion) was completed and red-black iteration was added for solving the HFEM-PN system with both preconditioned Jacobian-free Newton Krylov and Richardson iteration solvers. The HFEM-PN scheme, as one of the low-order transport schemes, is expected for supporting routine design simulations. Pin power reconstruction capability was also designed and implemented with the Griffin ISOXML module to enhance all the low-order transport solvers for more accurate multiphysics simulations. Numerical results are presented for demonstrating the capabilities and verifying their performance, and future works are discussed.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

EVs@Scale Next-Gen Profiles - EV Profile Capture 2023

As part of the U.S. DOE EVs@Scale consortium Next-Gen Profiles (NGP) project, the profile capture of production electric vehicles undergoing high power charging (HPC) is conducted over a wide range of conditions to explore variance and performance. Charge session parameters are collected from both the electric vehicle (EV) and electric vehicle supply equipment (EVSE) at a rate of 10Hz and entered into a time-series database for analysis. These charge profiles are captured under nominal and off-nominal conditions, exploring the impact of battery state of charge (SOC), battery temperature, vehicle condition, smart charge management (SCM), and EVSE limitations. Nominal conditions are defined to be ideal conditions that should transfer the maximum allowable energy in the minimum possible amount of time. Nominal condition profiles are compared across EVs to characterize state-of-the-art EV charging performance against one another. Off-nominal condition profiles are compared against its nominal condition profile counterpart to highlight the variance across less desirable starting conditions within a single EV. Under nominal conditions, most EVs can achieve the original equipment manufacturer (OEM) rated peak-power and charge times. Peak power across EVs has variance due to the vehicle design, battery topology, charging strategy, etc. These 10-100% nominal preconditioned charge profiles across 13 EVs (11 light-duty (LD), 2 heavy-duty (HD)) were used for analysis in power curve analysis, power distribution, SOC and range comparison, thermal impacts of current draw, battery pack size and energy charged, and ramp rates. Power curve analysis show the uniqueness in power vs time curves across all EVs, breaking down the features of a typical profile and where variation is typically seen. Power distribution results showed that over 50% of charge time is spent below 50kW and only 12.1% is spent above 200kW. SOC and range performance yielded different top performing EVs when exploring goals of performance from SOC and range gained after 10-min (EV2) and 20-min (EV8), and time-to-achieve 80% SOC (EV8) and 200 miles of range (EV1). Current draw from 400-volt EVs had a higher thermal impact to cable/connector temperatures when compared to 800-volt EVs, due to a higher current requirement to achieve similar power levels. There was high variance in C rating, a useful metric when comparing the relationship between peak/average charge session power and relative battery pack size, across EVs under test. Ramp rates during initial power transfer were examined, fastest and slowest speeds ranging from 192.5kW/second to 2.6kW/second respectively. A similar analysis of ramp rates was also conducted for OCPP curtailment testing, where EVs underwent a 2-minute 65A curtailment request before returning to full-power charge. Under off-nominal conditions, most EVs experienced variation from the nominal condition profiles. EVs that underwent the full set of NGP defined testing conditions were compared, analysis of which was categorized by 800-volt and 400-volt EVs. It should be noted that the 800-volt EVs had significantly higher peak power ratings than the 400-volt EVs, and thus were more prone to variance. Initial state of charge, battery temperature, and vehicle condition had a considerable impact on peak power levels and charge time for the 800-volt vehicles under test. EVSE limited tests lowered achievable power down to 200kW, greatly impacted 800-volt charging power, and had little to no effect on 400-volt charging power. Adapter testing was performed on a single 400-volt EV, where power and current limitations were found but overall charge time was not significantly impacted. Adapter and boost converter testing was performed on a single 800-volt EV, where both charge power and charge time were greatly impacted. Charge profiles are unique, and comparing such requires analysis that examines starting conditions, vehicle and battery topologies,

Charging↗
DOE OSTI2020

Sorbmatech{sup R}Cs, an Innovative Highly Selective Cs Sorbent for Wastewater Decontamination - 20065

The treatment of radioactive wastewater from the operating nuclear industry (energy production, reprocessing of spent fuel, decommissioning of shut-down facilities, etc.) or in case of incidental or accidental situation (such as Fukushima Daichii NPP accident) is an important issue requiring constant improvement of the technologies and techniques used to prevent radioactive releases into the environment and/or to reduce the production of secondary waste. CEA, Orano and CTI are collaborating since 2014, with the support of the French government's program 'investments for the future', to develop Sorbmatech{sup R}Cs, an innovative and patented selective sorbent (Grandjean and Barre 2014) to deal with the removal of radioactive {sup 137}Cs, one of the most hazardous isotope to be managed in nuclear waste, highly efficient either in agitated reactor or in column process. Sorbmatech{sup R}Cs sorbent consists of mesoporous silica particles loaded with potassium-copper hexacyanoferrate nanoparticles (KCu-HCF). This peculiar structure enables much faster sorption kinetics compared to denser crystalline structures, such as Silicotitanate or Zeolite sorbents, whose drawback consists in hindering the diffusion process through the material, and whose larger sorbent particles slow down the ion exchange process. The equilibrium of Sorbmatech{sup R}Cs is reached in few minutes, even at high flow rates in a fixed-bed column process (up to 10 m.h{sup -1} Darcy velocity) with low drop pressure. Moreover, Sorbmatech{sup R}Cs can be used over a wide pH range from 2 to 10 and has a very high selectivity towards {sup 137}Cs even in brine solution (saline, mineral or sea water). This new sorbent also has a K{sub d} higher than 10{sup 5} mL.g{sup -1} depending on the salinity of the wastewater, enabling a highly efficient removal of Cs trace in radioactive wastewater compared to Cs selective sorbent such as Zeolites and Silicotitanates. Its radiological resistance enables the treatment of high, medium and low activity wastewater. The CEA-Orano-CTI/ALSYS partnership leads to the development of a simple and low-cost synthetic route, which demonstrated large-scale manufacturing of the product on a batch of 100 kg. Two products have been developed and are ready for short run market introduction so far with grain size well adapted either for column processes or agitated reactors.. Finally, the industrial implementation has been studied and largely benefits from the properties of the sorbent. Indeed, its high sorption kinetics enable high flow rates resulting in a decrease of operation times, unit sizes, and volumes of secondary waste produced. Moreover, Sorbmatech{sup R}Cs is designed to be combined into a full and integrated solution, ranging from preliminary optimization of operating conditions through numerical dimensioning model, industrial implementation and treatment of the wastewater (through cartridge or preconditioned columns systems developed thanks to Orano's extensive expertise), to long-term stabilization of generated waste (under progress). (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2024

An Efficient Numerical Algorithm for Solving Coupled Time-Dependent Ginzburg-Landau Equation for Superconductivity and Elasticity

A decoupled finite element algorithm is developed for simulating the vortex dynamics on an elastic superconductor which couples the time-dependent Ginzburg- Landau equation with the complex-valued superconducting order parameter and the vector-valued magnetic potential, and the elasticity equation. We present an iterative algorithm for the decoupled system arising from the time and spatial discretization using a combination of preconditioner, algebraic multigrid method (AMG) and preconditioned conjugate gradient method (PCG). The iterative algorithm allows us to perform large-scale three-dimensional simulations of mesoscale pattern formation during superconducting phase transitions with arbitrary elastic boundary conditions. Here, the performance and efficiency of the algorithm are numerically verified by several benchmark problems, exhibiting up to two orders of magnitude improvement depending on the scale of discrete system compared to the exact solver.

Efficiency↗
DOE OSTI2020

A fully implicit, scalable, conservative nonlinear relativistic Fokker–Planck 0D-2P solver for runaway electrons

Upon application of a sufficiently strong electric field, electrons break away from thermal equilibrium and approach relativistic speeds. These highly energetic ‘runaway’ electrons (~ MeV) play a significant role in tokamak disruption physics, and therefore their accurate understanding is essential to develop reliable mitigation strategies. As such, we have developed a fully implicit solver for the 0D-2P (i.e., including two momenta coordinates) relativistic nonlinear Fokker–Planck equation (rFP). As in earlier implicit rFP studies (NORSE, CQL3D), electron–ion interactions are modeled using the Lorentz operator, and synchrotron damping using the Abraham–Lorentz–Dirac reaction term. However, our implementation improves on these earlier studies by (1) ensuring exact conservation properties for electron collisions, (2) strictly preserving positivity, and (3) being scalable algorithmically and in parallel. Key to our proposed approach is an efficient multigrid preconditioner for the linearized rFP equation, a multigrid elliptic solver for the Braams–Karney potentials, and a novel adaptive technique to determine the associated boundary values. We verify the accuracy and efficiency of the proposed scheme with numerical results ranging from small electric-field electrical conductivity measurements to the accurate reproduction of runaway tail dynamics when strong electric fields are applied.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2020

Approximate inverse-based block preconditioners in poroelasticity

We focus on the fully implicit solution of the linear systems arising from a three-field mixed finite element approximation of Biot’s poroleasticity equations. The objective is to develop algebraic block preconditioners for the efficient solution of such systems by Krylov subspace methods. In this work, we investigate the use of approximate inverse-based techniques to decouple the native system of equations and obtain explicit sparse approximations of the Schur complements related to the physics-based partitioning of the unknowns by field type. Here, the proposed methods are tested in various numerical experiments including real-world applications dealing with petroleum and geotechnical engineering.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Efficient solvers for hybridized three-field mixed finite element coupled poromechanics

We consider a mixed hybrid finite element formulation for coupled poromechanics. A stabilization strategy based on a macro-element approach is advanced to eliminate the spurious pressure modes appearing in undrained/incompressible conditions. The efficient solution of the stabilized mixed hybrid block system is addressed by developing a class of block triangular preconditioners based on a Schur-complement approximation strategy. Robustness, computational efficiency and scalability of the proposed approach are theoretically discussed and tested using challenging benchmark problems on massively parallel architectures.

42 ENGINEERING↗
DOE OSTI2025

Two-level overlapping additive Schwarz preconditioner for training scientific machine learning applications

In this work we introduce a novel two-level overlapping additive Schwarz preconditioner for accelerating the training of scientific machine learning applications. The design of the proposed preconditioner is motivated by the nonlinear two-level overlapping additive Schwarz preconditioner. The neural network parameters are decomposed into groups (subdomains) with overlapping regions. In addition, the network’s feed-forward structure is indirectly imposed through a novel subdomain-wise synchronization strategy and a coarse-level training step. Through a series of numerical experiments, which consider physicsinformed neural networks and operator learning approaches, we demonstrate that the proposed two-level preconditioner significantly speeds up the convergence of the standard (LBFGS) optimizer while also yielding more accurate machine learning models. Moreover, the devised preconditioner is designed to take advantage of model-parallel computations, which can further reduce the training time.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Multiscale formulation of frictional contact mechanics at the pore scale

Direct numerical simulation (DNS) yields the highest fidelity predictions of mechanical deformation at the pore scale, but is prohibitively expensive for analyzing large or many samples. Discrete element methods (DEM) are an efficient alternative, but are limited to granular media and incapable of estimating or controlling prediction errors. We present a pore-level multiscale method (PLMM) that approximates DNS efficiently and with controllable accuracy. We focus on the linear elastic response of a consolidated geologic porous medium with arbitrary microstructure, heterogeneous mineralogy, containing cracks or defects. PLMM decomposes the solid phase into non-overlapping subdomains, on which local basis functions are constructed. The bases are then coupled with a global interface problem that accounts for slip or stick contact conditions between the subdomains. PLMM produces an initial, but accurate, approximation to DNS that can be iteratively improved. It is amenable to parallelism and allows for different mesh, models, and physics in each subdomain. An algebraic interpretation of PLMM as a preconditioner is also presented to allow non-intrusive implementation into existing solvers. Lastly, this work extends previous developments of PLMM in fluid dynamics to solid mechanics and enables future extensions towards modeling coupled flow and mechanics problems.

42 ENGINEERING↗
DOE OSTI2023

A greedy Galerkin method to efficiently select sensors for linear dynamical systems

A key challenge in inverse problems is the selection of sensors to gather the most effective data. In this paper, we consider the problem of inferring the initial condition to a linear dynamical system and develop an efficient control-theoretical approach for greedily selecting sensors. Our method employs a Galerkin projection to reduce the size of the inverse problem, resulting in a computationally efficient algorithm for sensor selection. As a byproduct of our algorithm, we obtain a preconditioner for the inverse problem that enables the rapid recovery of the initial condition. Here, we analyze the theoretical performance of our greedy sensor selection algorithm as well as the performance of the associated preconditioner. Finally, we verify our theoretical results on various inverse problems involving partial differential equations.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Porting hypre to heterogeneous computer architectures: Strategies and experiences

We report that linear systems are occurring in many applications, and solving them can take a large amount of the total simulation time. The high performance library hypre provides a variety of interfaces and linear solvers, including various multigrid methods, that have achieved good scalability on a variety of homogeneous parallel computer architectures. Heterogeneous architectures with nodes that have both CPUs and accelerators provide new challenges, since they require more fine-grained parallelism and reduced data movement between different memories on a single node as well as across nodes. We will discuss our experiences and strategies to port hypre to heterogeneous computers with accelerators, including the design of a new memory model, the use of abstractions, the BoxLoop macros in the structured and semi-structured interfaces, and the restructuring of algebraic multigrid (AMG) into modular components. We present numerical experiments comparing CPU and GPU performance for several test problems.

97 MATHEMATICS AND COMPUTING↗