Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallel in Time”

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 1,063 records · Page 59
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Tula: Optimizing Time, Cost, and Generalization in Distributed Large-Batch Training

Distributed training increases the number of batches processed per iteration either by scaling-out (adding more nodes) or scaling-up (increasing the batch-size). However, the largest configuration does not necessarily yield the best performance. Horizontal scaling introduces additional communication overhead, while vertical scaling is constrained by computation cost and device memory limits. Thus, simply increasing the batch-size leads to diminishing returns: training time and cost decrease initially but eventually plateaus, creating a knee-point in the time/cost vs. batch-size pareto curve. The optimal batch-size therefore depends on the underlying model, data and available compute resources. Large batches also suffer from worse model quality due to the well-known “generalization gap”. In this paper, we present Tula, an online service that automatically optimizes time, cost, and convergence quality for large-batch training of convolutional models. It combines parallel-systems modeling with statistical performance prediction to identify the optimal batchsize. Tula predicts training time and cost within 7.5−14% error across multiple models, and achieves up to 20× overall speedup and improves test accuracy by ≈9% on average over standard large-batch training on various vision tasks, thus successfully mitigating the generalization gap and accelerating training at the same time.

Tyagi, Sahil [ORNL] (ORCID:0009000783144745)↗
DOE OSTI2025

High-throughput synthesis of high-entropy alloys via parallelized electric field assisted sintering

Materials discovery and design is an expensive and time-consuming process, though necessary to advance many engineering fields. In this work, a novel tooling design is utilized in conjunction with electric field assisted sintering (EFAS) to effectively create a new high-throughput synthesis technique: parallelized EFAS. Through this technique, a wide range of material compositions and geometries can be synthesized in parallel as isolated samples or as part of contiguous arrays. Multiple tooling designs are explored to examine both the flexibility and limitations of the technique. A series of increasing complex alloys is produced simultaneously using in situ alloying, beginning with pure Ni and adding equimolar constituents up to the septenary high-entropy alloy AlCoCrCuFeMnNi. Microstructural characterization reveals each sample is effectively fully dense and chemically homogenous while exhibiting phases in agreement with CALPHAD predictions. Scalability of parallelized EFAS is then experimentally demonstrated and the implications for materials discovery and automation are discussed.

36 - MATERIALS SCIENCE↗
NASA NTRS2002

Progress Toward Generation of a Navier-Stokes Database for a Harrier in Ground Effect

The Harrier YAV-8B aircraft is capable of vertical and short-field take-off and landing (V/STOL) by directing its four exhaust nozzles toward the ground, or conventional flight by rotating its nozzles into a horizontal position. The British Royal Air Force and the United States Marine Corps have used this aircraft for more than 30 years to provide a quick reaction time for troop support, and reduce the need for long runways. The success of this powered-lift (PL) vehicle has also prompted the more recent design of the Joint Strike Fighter (JSF). However there are significant safety issues that must be addressed when operating a PL vehicle in close proximity to the ground. Hot Gas Ingestion (HGI) by the inlets can result in a rapid loss of powered lift; and high-speed jet flows along the ground plane can induce low pressures underneath the vehicle, causing a 'suck-down' effect. Under these conditions, departure from controlled flight may occur. Moreover, unsteady ground vortices and jet fountains can affect the aircraft,s controllability and its proximity to ground troops. The viscous, time-dependent flow fields of PL vehicles are difficult to accurately and efficiently predict using Computational Fluid Dynamics (CFD). A number of researchers have used the time-dependent Reynolds-averaged Navier-Stokes (RANS) equations to compute flows for single and multiple jets in a cross-flow. A few have added some geometric complexity to the problem by computing flows for jet-augmented delta wings near a ground plane. Smith et.al. computed for the first time a single RANS solution about a simplified Harrier. This geometry included a fuselage, wing, leading edge root extension (LERX), inlets, and exhaust nozzles. All of these investigations cite two practical problems with computing these flows: 1) the need for improved solution accuracy; and, 2) the need for faster solution methods. We view the need for faster solution methods as key to improving the solution accuracy and making this class of computation more routine. One can hardly refine grids, explore the use of advanced turbulence models, and generate databases when it takes weeks of dedicated computer time for a single solution. Chaderjian, Ahmad, Pandya, and Murman have focused on reducing the time-to-solution for this very difficult and complex problem through process automation and exploitation of parallel computing. They began with the Harrier geometry reported, and added a deflected wing flap and empennage for greater realism. To date more than 80 solutions have been carried out. This paper will describe this process and progress made in reducing the time required to generate a simple longitudinal force and moment database for a Harrier in ground effect. It shows a typical snap-shot from an unsteady streakline animation, where fluid particles are colored by temperature. The ground vortex and a jet-fountain vortex are highlighted. It also shows a similar streakline image, where HGI occurs due to the vehicle in close proximity to the ground. It is show the mean lift coefficient as a function of angle of attack and height. The angle of attack range was 4 deg less than or = alpha less than or = 10 deg with an increment of 1 degree, and the height range was 10 ft less than or = h less than or = 30ft with an increment of 5 feet. This 35 solution database was extended to over 2500 cases using a monotone cubic-spline interpolation procedure. The suck-down effect (reduction of lift near the ground) is highlighted in the figure. The "cushion effect," the conventional reduction of lift as the vehicle moves out of ground effect, is also indicated. All 35 RANS solutions were obtained using 952 Silicon Graphics Origin 2000 and 3000 processors in dedicated mode for one week. Typically, 112 processors were assigned to each case. Some other cases used fewer processors to utilize all available CPUS. The final paper will report on the automation of the solution process, including: grid generation, job monitoring, solution completion criteria, and post processing. Moreover, improvements in parallel efficiency for a dual time-step algorithm for the RANS equations will also be presented. Results will be discussed in detail using unsteady streakline flow visualization to correlate unsteady flow structures with dominant aerodynamic frequencies. The stability derivatives, CL, and CL, will also be presented.

Chaderjian, Neal M.↗
DOE OSTI2016

SU-D-206-02: Evaluation of Partial Storage of the System Matrix for Cone Beam Computed Tomography Using a GPU Platform

Purpose: Iterative reconstruction algorithms in computed tomography (CT) require a fast method for computing the intersections between the photons’ trajectories and the object, also called ray-tracing or system matrix computation. This work evaluates different ways to store the system matrix, aiming to reconstruct dense image grids in reasonable time. Methods: We propose an optimized implementation of the Siddon’s algorithm using graphics processing units (GPUs) with a novel data storage scheme. The algorithm computes a part of the system matrix on demand, typically, for one projection angle. The proposed method was enhanced with accelerating options: storage of larger subsets of the system matrix, systematic reuse of data via geometric symmetries, an arithmetic-rich parallel code and code configuration via machine learning. It was tested on geometries mimicking a cone beam CT acquisition of a human head. To realistically assess the execution time, the ray-tracing routines were integrated into a regularized Poisson-based reconstruction algorithm. The proposed scheme was also compared to a different approach, where the system matrix is fully pre-computed and loaded at reconstruction time. Results: Fast ray-tracing of realistic acquisition geometries, which often lack spatial symmetry properties, was enabled via the proposed method. Ray-tracing interleaved with projection and backprojection operations required significant additional time. In most cases, ray-tracing was shown to use about 66 % of the total reconstruction time. In absolute terms, tracing times varied from 3.6 s to 7.5 min, depending on the problem size. The presence of geometrical symmetries allowed for non-negligible ray-tracing and reconstruction time reduction. Arithmetic-rich parallel code and machine learning permitted a modest reconstruction time reduction, in the order of 1 %. Conclusion: Partial system matrix storage permitted the reconstruction of higher 3D image grid sizes and larger projection datasets at the cost of additional time, when compared to the fully pre-computed approach. This work was supported in part by the Fonds de recherche du Quebec - Nature et technologies (FRQ-NT). The authors acknowledge partial support by the CREATE Medical Physics Research Training Network grant of the Natural Sciences and Engineering Research Council of Canada (Grant No. 432290).

60 APPLIED LIFE SCIENCES↗
DOE OSTI2025

1D model of SOL with self consistent calculation of non-coronalradiation and applications to shallow field line angles

A 1D model of the scrape off layer (SOL) is created with a physics based calculation of the impurity radiation and the inclusion of the perpendicular heat flux q ⊥ compared to previous 1D models. The calculation of the impurity radiation includes the addition of a model using parallel and perpendicular motion to calculate the impurity confinement time τ and the inclusion of charge exchange between neutral hydrogen and impurity ions. Reasonable agreement is found between the model and the 2D fluid code SOLPS-ITER. Additionally, an improvement is shown over the constant τ used in previous works The model is used to examine the effects of shallow field line angles at ITER levels of parallel heat flux (q || ). A large increase in radiation results in an increase in the upstream parallel heat flux which can still be in detachment at shallow field line angles for both q ⊥ and the calculated τ for traditional impurities (argon and nitrogen) as well as boron. The physics behind this increase in radiation at shallow field line angles is shown to be a decrease in τ and a change in the temperature profile with the inclusion of q ⊥ .

ITER↗
DOE OSTI2014

Development of Variational Guiding Center Algorithms for Parallel Calculations in Experimental Magnetic Equilibria

Structure-preserving algorithms obtained via discrete variational principles exhibit strong promise for the calculation of guiding center test particle trajectories. The non-canonical Hamiltonian structure of the guiding center equations forms a novel and challenging context for geometric integration. To demonstrate the practical relevance of these methods, a prototypical variational midpoint algorithm is applied to an experimental magnetic equilibrium. The stability characteristics, conservation properties, and implementation requirements associated with the variational algorithms are addressed. Furthermore, computational run time is reduced for large numbers of particles by parallelizing the calculation on GPU hardware.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
NASA NTRS1994

Automated Instrumentation, Monitoring and Visualization of PVM Programs Using AIMS

We present views and analysis of the execution of several PVM codes for Computational Fluid Dynamics on a network of Sparcstations, including (a) NAS Parallel benchmarks CG and MG (White, Alund and Sunderam 1993); (b) a multi-partitioning algorithm for NAS Parallel Benchmark SP (Wijngaart 1993); and (c) an overset grid flowsolver (Smith 1993). These views and analysis were obtained using our Automated Instrumentation and Monitoring System (AIMS) version 3.0, a toolkit for debugging the performance of PVM programs. We will describe the architecture, operation and application of AIMS. The AIMS toolkit contains (a) Xinstrument, which can automatically instrument various computational and communication constructs in message-passing parallel programs; (b) Monitor, a library of run-time trace-collection routines; (c) VK (Visual Kernel), an execution-animation tool with source-code clickback; and (d) Tally, a tool for statistical analysis of execution profiles. Currently, Xinstrument can handle C and Fortran77 programs using PVM 3.2.x; Monitor has been implemented and tested on Sun 4 systems running SunOS 4.1.2; and VK uses X11R5 and Motif 1.2. Data and views obtained using AIMS clearly illustrate several characteristic features of executing parallel programs on networked workstations: (a) the impact of long message latencies; (b) the impact of multiprogramming overheads and associated load imbalance; (c) cache and virtual-memory effects; and (4significant skews between workstation clocks. Interestingly, AIMS can compensate for constant skew (zero drift) by calibrating the skew between a parent and its spawned children. In addition, AIMS' skew-compensation algorithm can adjust timestamps in a way that eliminates physically impossible communications (e.g., messages going backwards in time). Our current efforts are directed toward creating new views to explain the observed performance of PVM programs. Some of the features planned for the near future include: (a) ConfigView, showing the physical topology of the virtual machine, inferred using specially formatted IP (Internet Protocol) packets; and (b) LoadView, synchronous animation of PVM-program execution and resource-utilization patterns.

Mehra, Pankaj↗
DOE OSTI2021

PLANC: Parallel Low-rank Approximation with Nonnegativity Constraints

In this work, we consider the problem of low-rank approximation of massive dense nonnegative tensor data, for example, to discover latent patterns in video and imaging applications. As the size of data sets grows, single workstations are hitting bottlenecks in both computation time and available memory. We propose a distributed-memory parallel computing solution to handle massive data sets, loading the input data across the memories of multiple nodes, and performing efficient and scalable parallel algorithms to compute the low-rank approximation. We present a software package called Parallel Low-rank Approximation with Nonnegativity Constraints, which implements our solution and allows for extension in terms of data (dense or sparse, matrices or tensors of any order), algorithm (e.g., from multiplicative updating techniques to alternating direction method of multipliers), and architecture (we exploit GPUs to accelerate the computation in this work). We describe our parallel distributions and algorithms, which are careful to avoid unnecessary communication and computation, show how to extend the software to include new algorithms and/or constraints, and report efficiency and scalability results for both synthetic and real-world data sets.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1989

The Real Time Interactive Display Environment (RTIDE), a display building tool developed by Space Shuttle flight controllers

NASA's Mission Control Center, located at Johnson Space Center, is incrementally moving from a centralized architecture to a distributed architecture. Starting with STS-29, some host-driven console screens will be replaced with graphics terminals driven by workstations. These workstations will be supplied realtime data first by the Real Time Data System (RTDS), a system developed inhouse, and then months later (in parallel with RTDS) by interim and subsequently operational versions of the Mission Control Center Upgrade (MCCU) software package. The Real Time Interactive Display Environment (RTIDE) was built by Space Shuttle flight controllers to support the rapid development of multiple new displays to support Shuttle flights. RTIDE is a display building tool that allows non-programmers to define object-oriented, event-driven, mouseable displays. Particular emphasis was placed on upward compatibility between RTIDE versions, ability to acquire data from different data sources, realtime performance, ability to modularly upgrade RTIDE, machine portability, and a clean, powerful user interface. The operational and organizational factors that drove RTIDE to its present form, the actual design itself, simulation and flight performance, and lessons learned in the process are discussed.

Kalvelage, Thomas A.↗
DOE OSTI2024

SPARC-X: Quantum simulations at extreme scale - reactive dynamics from first principles

We have developed the massively parallel electronic structure code SPARC-X: a computational framework for performing Kohn-Sham Density Functional Theory (DFT) calculations that can scale linearly with the number of atoms in the system, while being able to leverage petascale and emerging exascale parallel computers to study chemical phenomena at unprecedented length and time scales. SPARC-X exploits a recent breakthrough in electronic structure methodologies: systematically improvable, strictly local, orthonormal, discontinuous real-space bases that efficiently and systematically capture the local chemistry of the system. With further adaptation using new machine-learning techniques and the use of the massively parallel Spectral Quadrature (SQ) electronic structure method, the algorithmic complexity and prefactor associated with DFT calculations involving semilocal as well as hybrid functionals are dramatically reduced. Using petascale computational resources, SPARC-X enables quantum mechanical simulations at length and time scales previously accessible only by empirical approaches, e.g., 1,000,000 atoms for a few picoseconds using semilocal functionals or 1,000 atoms for a few picoseconds using hybrid functionals. Using exascale resources, the sizes and times targeted are two orders of magnitude larger. Such a capability has applications in a wide variety of chemical sciences, including reactive interfaces where large length- and/or long time-scales are needed and traditional force fields fail. This is particularly important in dynamic catalysis, where bond breaking and formation must be understood in detail. We developed, tested, and employed the SPARC-X framework to understand the photocatalytic properties of TiO 2 nanoparticles, revealing finite size effects that cannot be captured with standard model systems or functionals. This integrated development and application strategy ensures that SPARC-X remains a robust, efficient, and scalable software package for quantum simulations on current petascale and emerging exascale computing resources.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2015

A Heuristic Approach to Correlating ERAM Flight Data from Twenty Centers

Among its many other functions, the Federal Aviation Administration’s En Route Automation Modernization (ERAM) provides external systems with real-time air traffic data for flights in enroute airspace in the National Airspace System. It replaced the En Route Host computer and backup system used at 20 FAA Air Route Traffic Control Centers (Centers) nationwide. Among the new features of ERAM, its output data stream of flight plan and track data includes a unique identifier for a flight originating in any one of the 20 ERAM Centers. The unique identifier, called the Global Unique Flight Identifier (GUFI), is persistent across all the Centers that track the flight. However, certain factors make it difficult to correlate data using the GUFI. First, the value of the GUFI is only unique within a time window of seven days. Second, the GUFI is attached only to flight-plan related data messages. Finally, track positions reported by ERAM do not reference the GUFI. In order to correlate historical as well as real time flight-plan and position related ERAM data, an efficient, heuristic approach was developed, and a prototype was developed. The approach showed that the processing speed, through parallel processing, is sufficient to correlate ERAM data in real-time. As described in this paper, when there are multiple track positions reported from multiple Centers within a few seconds, each position is assigned with a weighted score to indicate the quality of the position relative to its last know position. The weighted score can be used to eliminate potentially duplicate track positions. The approach is database-agnostic, and can be implemented in a Big Data system such as an Apache Hadoop system, as well as in traditional database systems.

Correlating ERAM Flight Data↗
DOE OSTI2014

A high-order discontinuous Galerkin method for fluid–structure interaction with efficient implicit–explicit time stepping

We present a high-order accurate scheme for coupled fluid–structure interaction problems. The fluid is discretized using a discontinuous Galerkin method on unstructured tetrahedral meshes, and the structure uses a high-order volumetric continuous Galerkin finite element method. Standard radial basis functions are used for the mesh deformation. The time integration is performed using a partitioned approach based on implicit–explicit Runge–Kutta methods. The resulting scheme fully decouples the implicit solution procedures for the fluid and the solid parts, which we perform using two separate efficient parallel solvers. We demonstrate up to fifth order accuracy in time on a non-trivial test problem, on which we also show that additional subiterations are not required. We solve a benchmark problem of a cantilever beam in a shedding flow, and show good agreement with other results in the literature. Finally, we solve for the flow around a thin membrane at a high angle of attack in both 2D and 3D, and compare with the results obtained with a rigid plate.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
NASA NTRS1994

Design and implementation of a parallel unstructured Euler solver using software primitives

This paper is concerned with the implementation of a three-dimensional unstructured-grid Euler solver on massively parallel distributed-memory computer architectures. The goal is to minimize solution time by achieving high computational rates with a numerically efficient algorithm. An unstructured multigrid algorithm with an edge-based data structure has been adopted, and a number of optimizations have been devised and implemented to accelerate the parallel computational rates. The implementation is carried out by creating a set of software tools, which provide an interface between the parallelization issues and the sequential code, while providing a basis for future automatic run-time compilation support. Large practical unstructured grid problems are solved on the Intel iPSC/860 hypercube and Intel Touchstone Delta machine. The quantitative effects of the various optimizations are demonstrated, and we show that the combined effect of these optimizations leads to roughly a factor of 3 performance improvement. The overall solution efficiency is compared with that obtained on the Cray Y-MP vector supercomputer.

Das, R.↗
NASA NTRS1994

Analysis of DE-1 PWI electric field data

The measurement of low frequency electric field oscillations may be accomplished with the Plasma Wave Instrument (PWI) on DE 1. Oscillations at a frequency around 1 Hz are below the range of the conventional plasma wave receivers, but they can be detected by using a special processing of the quasi-static electric field data. With this processing it is also possible to determine if the electric field oscillations are predominately parallel or perpendicular to the ambient magnetic field. The quasi-static electric field in the DE 1 spin/orbit plane is measured with a long-wire 'double probe'. This antenna is perpendicular to the satellite spin axis, which in turn is approximately perpendicular to the geomagnetic field in the polar magnetosphere. The electric field data are digitally sampled at a frequency of 16 Hz. The measured electric field signal, which has had phase reversals introduced by the rotating antenna, is multiplied by the sine of the rotation angle between the antenna and the magnetic field. This is called the 'perpendicular' signal. The measured time series is also multiplied with the cosine of the angle to produce a separate 'parallel' signal. These two separate time series are then processed to determine the frequency power spectrum.

Weimer, Daniel↗
NASA NTRS1998

I/O Parallelization for the Goddard Earth Observing System Data Assimilation System (GEOS DAS)

The National Aeronautics and Space Administration (NASA) Data Assimilation Office (DAO) at the Goddard Space Flight Center (GSFC) has developed the GEOS DAS, a data assimilation system that provides production support for NASA missions and will support NASA's Earth Observing System (EOS) in the coming years. The GEOS DAS will be used to provide background fields of meteorological quantities to EOS satellite instrument teams for use in their data algorithms as well as providing assimilated data sets for climate studies on decadal time scales. The DAO has been involved in prototyping parallel implementations of the GEOS DAS for a number of years and is now embarking on an effort to convert the production version from shared-memory parallelism to distributed-memory parallelism using the portable Message-Passing Interface (MPI). The GEOS DAS consists of two main components, an atmospheric General Circulation Model (GCM) and a Physical-space Statistical Analysis System (PSAS). The GCM operates on data that are stored on a regular grid while PSAS works with observational data that are scattered irregularly throughout the atmosphere. As a result, the two components have different data decompositions. The GCM is decomposed horizontally as a checkerboard with all vertical levels of each box existing on the same processing element(PE). The dynamical core of the GCM can also operate on a rotated grid, which requires communication-intensive grid transformations during GCM integration. PSAS groups observations on PEs in a more irregular and dynamic fashion.

Lucchesi, Rob↗
NASA NTRS1988

Parallel algorithm for determining motion vectors in ice floe images by matching edge features

A parallel algorithm is described to determine motion vectors of ice floes using time sequences of images of the Arctic ocean obtained from the Synthetic Aperture Radar (SAR) instrument flown on-board the SEASAT spacecraft. Researchers describe a parallel algorithm which is implemented on the MPP for locating corresponding objects based on their translationally and rotationally invariant features. The algorithm first approximates the edges in the images by polygons or sets of connected straight-line segments. Each such edge structure is then reduced to a seed point. Associated with each seed point are the descriptions (lengths, orientations and sequence numbers) of the lines constituting the corresponding edge structure. A parallel matching algorithm is used to match packed arrays of such descriptions to identify corresponding seed points in the two images. The matching algorithm is designed such that fragmentation and merging of ice floes are taken into account by accepting partial matches. The technique has been demonstrated to work on synthetic test patterns and real image pairs from SEASAT in times ranging from .5 to 0.7 seconds for 128 x 128 images.

Manohar, M.↗
DOE OSTI2018

Alternating Anderson–Richardson method: An efficient alternative to preconditioned Krylov methods for large, sparse linear systems

Here, we present the Alternating Anderson–Richardson (AAR) method: an efficient and scalable alternative to preconditioned Krylov solvers for the solution of large, sparse linear systems on high performance computing platforms. Specifically, we generalize the recently proposed Alternating Anderson–Jacobi (AAJ) method (Pratapa et al., 2016) to include preconditioning, discuss efficient parallel implementation, and provide serial MATLAB and parallel C/C++ implementations. In serial applications to nonsymmetric systems, we find that AAR is comparably robust to GMRES, using the same preconditioning, while often outperforming it in time to solution; and find AAR to be more robust than Bi-CGSTAB for the problems considered. In parallel applications to the Helmholtz and Poisson equations, we find that AAR shows superior strong and weak scaling to GMRES, Bi-CGSTAB, and Conjugate Gradient (CG) methods, using the same preconditioning, with consistently shorter times to solution at larger processor counts. Finally, in massively parallel applications to the Poisson equation, on up to 110,592 processors, we find that AAR shows superior strong and weak scaling to CG, with shorter minimum time to solution. We thus find that AAR offers a robust and efficient alternative to current state-of-the-art solvers, with increasing advantages as the number of processors grows.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

Deploy threading in Nalu solver stack

The goal of the ExaWind project is to enable predictive simulations of wind farms composed of many MW-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines, and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources. The primary code in the ExaWind project is Nalu, which is an unstructured-grid solver for the acousticallyincompressible Navier-Stokes equations, and mass continuity is maintained through pressure projection. The model consists of the mass-continuity Poisson-type equation for pressure and a momentum equation for the velocity. For such modeling approaches, simulation times are dominated by linear-system setup and solution for the continuity and momentum systems. For the ExaWind challenge problem, the moving meshes greatly affect overall solver costs as re-initialization of matrices and re-computation of preconditioners is required at every time step In this Milestone, we examine the effect of threading on the solver stack performance against flat-MPI results obtained from previous milestones using Haswell performance data full-turbine simulations. Whereas the momentum equations are solved only with the Trilinos solvers, we investigate two algebraic-multigrid preconditioners for the continuity equations: Trilinos/Muelu and HYPRE/BoomerAMG. These two packages embody smoothed-aggregation and classical Ruge-Stiiben AMG methods, respectively. In our FY18 Q2 report, we described our efforts to improve setup and solve of the continuity equations under flat-MPI parallelism. While significant improvement was demonstrated in the solve phase, setup times remained larger than expected. Starting with the optimized settings described in the Q2 report, we explore here simulation performance where OpenMP threading is employed in the solver stack. For Trilinos, threading is acheived through the Kokkos abstraction where, whereas HYPRE/BoomerAMG employs straight OpenMP. We examined results for our mid-resolution baseline turbine simulation configuration (229M DOF). Simulations on 2048 Haswell cores explored the effect of decreasing the number of MPI ranks while increasing the number of threads. Both HYPRE and Trilinos exhibited similar overal solution times, and both showed dramatic increases in simulation time in the shift from MPI ranks to OpenMP threads. This increase is attributed to the large amount of work per MPI rank starting at the single-thread configuration. Decreasing MPI ranks, while increasing threads, may be increasing simulation time due to thread synchronization and start-up overhead contributing to the latency and serial time in the model. These result showed that an MPI+OpenMP parallel decomposition will be more effective as the amount per MPI rank computation per MPI rank decreases and the communication latency increases. This idea was demonstrated in a strong scaling study of our low-resolution baseline model (29M DOF) with the Trilinos-HYPRE configuration. While MPI-only results showed scaling improvement out to about 1536 cores, engaging threading carried scaling improvements out to 4128 cores — roughly 7000 DOF per core. This is an important result as improved strong scaling is needed for simulations to be executed over sufficiently long simulated durations (i.e., for many timesteps). In addition to threading work described above, the team examined solver-performance improvements by exploring communication-overhead in the HYPRE-GMRES implementation through a communicationoptimal- GMRE algorithm (CO-GMRES), and offloading compute-intensive solver actions to GPUs. To those ends, a HYPRE mini-app was allow us to easily test different solver approaches and HYPRE parameter settings without running the entire Nalu code. With GPU acceleration on the Summitdev supercomputer, a 20x speedup was achieved for the overall preconditioner and solver execution time for the mini-app. A study on Haswell processors showed that CO-GMRES provides benefits as one increases MPI ranks.

17 WIND ENERGY↗