Search NASA⌕ Search

SEARCH · Search NASA

Results for “linear solver”

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

Mapping Spiking Neural Networks to Heterogeneous Crossbar Architectures using Integer Linear Programming

Advances in novel hardware devices and architectures allow Spiking Neural Network (SNN) evaluation using ultra-low power, mixed-signal, memristor crossbar arrays. As individual network sizes quickly scale beyond the dimensional capabilities of single crossbars, networks must be mapped onto multiple crossbars. Crossbar sizes within modern Memristor Crossbar Architectures (MCAs) are determined predominately not by device technology but by network topology; more, smaller crossbars consume less area thanks to the high structural sparsity found in larger, brain-inspired SNNs. Motivated by continuing increases in SNN sparsity due to improvements in training methods, we propose utilizing heterogeneous crossbar sizes to further reduce area consumption. This approach was previously unachievable as prior compiler studies only explored solutions targeting homogeneous MCAs. Our work improves on the state-of-the-art by providing Integer Linear Programming (ILP) formulations supporting arbitrarily heterogeneous architectures. By modeling axonal interactions between neurons, our methods produce better mappings while removing inhibitive a priori knowledge requirements. We first show a 16.7-27.6% reduction in area consumption for square-crossbar homogeneous architectures. Then, we demonstrate 66.9-72.7% further reduction when using a reasonable configuration of heterogeneous crossbar dimensions. Next, we present a new optimization formulation capable of minimizing the number of inter-crossbar routes. When applied to solutions already near-optimal in area, an 11.9-26.4% routing reduction is observed without impacting area consumption. Finally, we present a profile-guided optimization capable of minimizing the number of runtime spikes between crossbars. Compared to the best-area-then-route optimized solutions, we observe a further 0.5-14.8% inter-crossbar spike reduction while requiring 1–3 orders of magnitude less solver time.

Pohl, Devin [ORNL] (ORCID:0009000040149027)↗

Center for Extended Magnetohydrodynamics Modeling - Final Technical Report

This project funding supported approximately 74 percent of a Ph.D. graduate student, not including costs of travel and supplies. We had a highly successful research project including the development of a second-order implicit electromagnetic kinetic ion hybrid model [Cheng 2013, Sturdevant 2016], direct comparisons with the extended MHD NIMROD code and kinetic simulation [Schnack 2013], modeling of slab tearing modes using the fully kinetic ion hybrid model and finally, modeling global tearing modes in cylindrical geometry using gyrokinetic simulation [Chen 2015, Chen 2016]. We developed an electromagnetic second-order implicit kinetic ion fluid electron hybrid model [Cheng 2013]. As a first step, we assumed isothermal electrons, but have included drift-kinetic electrons in similar models [Chen 2011]. We used this simulation to study the nonlinear evolution of the tearing mode in slab geometry, including nonlinear evolution and saturation [Cheng 2013]. Later, we compared this model directly to extended MHD calculations using the NIMROD code [Schnack 2013]. In this study, we investigated the ion-temperature-gradient instability with an extended MHD code for the first time and got reasonable agreement with the kinetic calculation in terms of linear frequency, growth rate and mode structure. We then extended this model to include orbit averaging and sub-cycling of the ions and compared directly to gyrokinetic theory [Sturdevant 2016]. This work was highlighted in an Invited Talk at the International Conference on the Numerical Simulation of Plasmas in 2015. The orbit averaging sub-cycling multi-scale algorithm is amenable to hybrid architectures with GPUS or math co-processors. Additionally, our participation in the Center for Extend Magnetohydrodynamics motivated our research on developing the capability for gyrokinetic simulation to model a global tearing mode. We did this in cylindrical geometry where the results could be benchmarked with existing eigenmode calculations. First, we developed a gyrokinetic code capable of simulating long wavelengths using a fluid electron model [Chen 2015]. We benchmarked this code with an eigenmode calculation. Besides having to rewrite the field solver due to the breakdown in the gyrokinetic ordering for long wavelengths, very high radial resolution was required. We developed a technique where we used the solution from the eigenmode solver to specify radial boundary conditions allowing for a very high radial resolution of the inner solution. Using this technique enabled us to use our direct algorithm with gyrokinetic ions and drift kinetic electrons [Chen 2016]. This work was highlighted in an Invited Talk at the American Physical Society - Division of Plasma Physics in 2015.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Rythmos: Solution and Analysis Package for Differential-Algebraic and Ordinary-Differential Equations

Time integration is a central component for most transient simulations. It coordinates many of the major parts of a simulation together, e.g., a residual calculation with a transient solver, solution with the output, various operator-split physics, and forward and adjoint solutions for inversion. Even though there is this variety in these transient simulations, there is still a common set of algorithms and procedures to progress transient solutions for ordinary-differential equations (ODEs) and differential-alegbraic equations (DAEs). Rythmos is a collection of these algorithms that can be used for the solution of transient simulations. It provides common time-integration methods, such as Backward and Forward Euler, Explicit and Implicit Runge-Kutta, and Backward-Difference Formulas. It can also provide sensitivities, and adjoint components for transient simulations. Rythmos is a package within Trilinos, and requires some other packages (e.g., Teuchos and Thrya) to provide basic time-integration capabilities. It also can be coupled with several other Trilinos packages to provide additional capabilities (e.g., AztecOO and Belos for linear solutions, and NOX for non-linear solutions). The documentation is broken down into three parts: Theory Manual, User's Manual, and Developer's Guide. The Theory Manual contains the basic theory of the time integrators, the nomenclature and mathematical structure utilized within Rythmos, and verification results demonstrating that the designed order of accuracy is achieved. The User's Manual provides information on how to use the Rythmos, description of input parameters through Teuchos Parameter Lists, and description of convergence test examples. The Developer's Guide is a high-level discussion of the design and structure of Rythmos to provide information to developers for the continued development of capabilities. Details of individual components can be found in the Doxygen webpages.

97 MATHEMATICS AND COMPUTING↗

Computational Study of Droplet Dynamics in Dispersed Flow Film Boiling Regime in PWR Geometries

Dispersed flow film boiling (DFFB) regime is exhibited during low-flooding rate conditions following the postulated Loss-of-Coolant Accident (LOCA) in nuclear reactor flow channels. LOCA is a type of design-basis accident scenario, which is crucial for reactor design and licensing. Accurate models to predict the heat transfer in DFFB regime is critical for the reactor to maintain safety limits. The nominal operating linear heat rate is also set based on the maximum linear heat rate determined for a LOCA transient. It is understood that the behavior of entrained droplets in DFFB regime greatly affects the heat transfer by evaporating and desuperheating the surrounding steam. Furthermore, interaction with flow structures also aids in increasing the evaporation rate of entrained droplets. In order to quantify this enhancement in heat transfer, the size of droplets downstream of flow structures (Spacer Grids and Mixing Vanes) is of crucial importance. Thermal hydraulic analysis codes like COBRA-TF, CATHARE-3 and adopt a three-field equation model that allows them to model the droplet field separately. However, to ensure the accuracy of these field equations, various correlations and physical models are required. This paper proposes a method for characterizing the mean Sauter Diameter of droplets across spacer-grids and mixing vanes in PWR sub-channels using PHASTA - a high fidelity flow solver with multiphase capabilities using the level-set method. The PHASTA code is a finite element based computational fluid dynamics code designed for solving compressible and incompressible turbulent flows. It has been demonstrated to be massively scalable (up to 786,432 cores), which is crucial for simulating large realistic domains. The code incorporates the level-set method for interface tracking of the gas-liquid interface.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Counterflow diffusion flames of general fluids: Oxygen/hydrogen mixtures

A comprehensive framework has been established for studying laminar counterflow diffusion flames for general fluids over the entire regime of thermodynamic states. The model incorporates a unified treatment of fundamental thermodynamic and transport theories into an existing flow solver DMCF to treat detailed chemical kinetic mechanisms and multispecies transport. The resultant scheme can thus be applied to fluids in any state. Both subcritical and supercritical conditions are considered. As a specific example, diluted and undiluted H{sub 2}/O{sub 2} flames are investigated at pressures of 1-25 MPa and oxygen inlet temperatures of 100 and 300 K. The effects of pressure p and strain rate {epsilon}{sub s} on the heat release rate q{sub s}-dot, extinction limit, and flame structure are examined. In addition, the impact of cross-diffusion terms, such as the Soret and Dufour effects, on the flame behavior is assessed. Results indicate that the flame thickness {delta}{sub f} and heat release rate correlate well with the square root of the pressure multiplied by the strain rate. The strain rate at the extinction limit exhibits a quasi-linear dependence on p. Significant real-fluid effects take place in the transcritical regimes, as evidenced by the steep property variations in the local flowfield. However, their net influence on the flame properties appears to be limited due to the ideal-gas behavior of fluids in the high-temperature zone. (author)

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An Efficient Multiscale Finite-Element Method for Frequency-Domain Seismic Wave Propagation

The frequency-domain seismic-wave equation, that is, the Helmholtz equation, has many important applications in seismological studies, yet is very challenging to solve, particularly for large geological models. Iterative solvers, domain decomposition, or parallel strategies can partially alleviate the computational burden, but these approaches may still encounter nontrivial difficulties in complex geological models where a sufficiently fine mesh is required to represent the fine-scale heterogeneities. We develop a novel numerical method to solve the frequency-domain acoustic wave equation on the basis of the multiscale finite-element theory. We discretize a heterogeneous model with a coarse mesh and employ carefully constructed high-order multiscale basis functions to form the basis space for the coarse mesh. Solved from medium- and frequency-dependent local problems, these multiscale basis functions can effectively capture themedium’s fine-scale heterogeneity and the source’s frequency information, leading to a discrete system matrix with a much smaller dimension compared with those from conventional methods.We then obtain an accurate solution to the acoustic Helmholtz equation by solving only a small linear system instead of a large linear system constructed on the fine mesh in conventional methods.We verify our new method using several models of complicated heterogeneities, and the results show that our new multiscale method can solve the Helmholtz equation in complex models with high accuracy and extremely low computational costs.

58 GEOSCIENCES↗

Impact of Vibrational Nonequilibrium on the Supersonic Mode in Hypersonic Boundary Layers

The supersonic mode in hypersonic boundary layers has recently been shown to achieve a larger peak growth rate than the traditional second mode in particular circumstances relevant to free flight. In this environment, high levels of aerodynamic heating are encountered such that incorporation of chemical and vibrational nonequilibrium effects can become critical. The impact of thermal nonequilibrium (TNE) on the supersonic mode in particular has not been thoroughly and systematically investigated. This work uses thermochemical nonequilibrium (TCNE) direct numerical simulation (DNS) and TCNE linear stability theory (LST) to obtain a more complete investigation of the supersonic mode using both nonequilibrium and frozen gas models. The simulation is Mach 5 flow over a 1-mm-nose-radius axisymmetric cone 1 m in length, representative of experimental conditions in ground test facilities. Both LST and DNS results indicate that TNE effects are destabilizing compared with frozen flow and can lead to an N-factor difference of approximately +1 at the end of the cone. LST results also indicate that TNE effects are stabilizing to the second and supersonic modes in comparison with thermal equilibrium. The supersonic mode appeared to be slightly more affected by TNE than the second mode. In conclusion, collectively, the findings in this work suggest that even in ground test facilities where TCNE effects are relatively small compared with free flight, incorporating nonequilibrium models in both the mean flow and stability solvers can have a practical effect on hypersonic boundary-layer transition prediction analyses.

42 ENGINEERING↗

Investigation of the possibility to use a fine-mesh solver for resolving coupled neutronics and thermal-hydraulics

The development of a fine-mesh coupled neutronic/thermal-hydraulic solver is touched upon in this paper. The reported work investigates the feasibility of using finite volume techniques to discretize a set of conservation equations modeling neutron transport, fluid dynamics, and heat transfer within a single numerical tool. With the long-term objective of developing fine-mesh computing capabilities for a few selected fuel assemblies in a nuclear core, this preliminary study considers an infinite array of a single fuel assembly having a finite height. Thermal-hydraulic conditions close to the ones existing in PWRs are taken as a first test case. The neutronic modeling relies on the diffusion approximation in a multi-energy group formalism, with cross-sections pre-calculated and tabulated at the sub-pin level using a Monte Carlo technique. The thermal-hydraulics is based on the Navier-Stokes equations, complemented by an energy conservation equation. The non-linear coupling terms between the different conservation equations are fully resolved using classical iteration techniques. Early tests demonstrate that the numerical tool provides an unprecedented level of details of the coupled solution estimated within the same numerical tool and thus avoiding any external data transfer, using fully consistent models between the neutronics and the thermal-hydraulics. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Probability of Initiation in Neutron Transport

We discuss the numerical solution of the nonlinear integro-differential equation for the probability of a divergent neutron chain in a stationary system (i.e., the probability of initiation (POI)). We follow the development described in Bell’s classic paper on the stochastic theory of neutron transport. As noted by Bell, the linearized form of this equation resembles the linear adjoint neutron transport equation. A matrix formalism for the discretized steady state (or forward) neutron equation in slab geometry is first developed, and is then used to derive the discrete adjoint equation. A main advantage of this discrete development is that the resulting discrete adjoint equation does not depend upon how the multigroup cross sections for the forward problem are obtained. That is, we derive the discrete adjoint directly from the discrete forward equations rather than discretizing directly the adjoint equation. This also guarantees that the discrete adjoint operator is consistent with the inner product used to define the adjoint operator. We discuss three approaches for the numerical solution of the POI equations, and present numerical results on several test problems. The three solution methods are a simple fixed point iteration, a second approach that is akin to a nonlinear Power iteration, and a third approach which uses a Newton-Krylov nonlinear solver. We also give sufficient conditions to guarantee the existence and uniqueness of nontrivial solutions to our discrete POI equations when the discrete system is supercritical, and that only the trivial solution exists when the discrete system is subcritical. Our approach is modelled after the analysis presented for the continuous POI equations by Mokhtar-Kharroubi and Jarmouni-Idrissi [12], and by Pazy and Rabinowitz [14].

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Probability of Initiation in Neutron Transport

We discuss the numerical solution of the nonlinear integro-differential equation for the probability of a divergent neutron chain in a stationary system (i.e., the probability of initiation (POI)). We follow the development described in Bell’s classic paper on the stochastic theory of neutron transport. As noted by Bell, the linearized form of this equation resembles the linear adjoint neutron transport equation. A matrix formalism for the discretized steady state (or forward) neutron equation in slab geometry is first developed and is then used to derive the discrete adjoint equation. A main advantage of this discrete development is that the resulting discrete adjoint equation does not depend upon how the multigroup cross sections for the forward problem are obtained. That is, we derive the discrete adjoint directly from the discrete forward equations rather than discretizing directly the adjoint equation. This also guarantees that the discrete adjoint operator is consistent with the inner product used to define the adjoint operator. We discuss three approaches for the numerical solution of the POI equations, and present numerical results on several test problems. The three solution methods are a simple fixed-point iteration, a second approach that is akin to a nonlinear Power iteration, and a third approach which uses a Newton-Krylov nonlinear solver. We also give sufficient conditions to guarantee the existence and uniqueness of nontrivial solutions to our discrete POI equations when the discrete system is supercritical, and that only the trivial solution exists when the discrete system is subcritical. Our approach is modeled after the analysis presented for the continuous POI equations by Mokhtar-Kharroubi and Jarmouni-Idrissi, and by Pazy and Rabinowitz.

42 ENGINEERING↗

Simulations of the churning mode: Toroidally symmetric plasma convection and turbulence around the X-points in a snowflake divertor

Using a reduced MHD model, extended to include field-aligned thermal conduction, we present numerical simulations of the churning mode (CM): a toroidally symmetric, non-linear plasma vortex in the vicinity of the null points in a snowflake (SF) divertor [D. D. Ryutov et al., Phys. Scr. 89, 088002 (2014)]. Simulations are carried out across a range of inter-null separations, d xx ⁠, and inter-null orientations, θ⁠, primarily in conditions relevant to the MAST-U tokamak. We find that, when d xx is small, the CM induces additional transport across the X-points when β pm ≳8%, where β pm is the ratio of the plasma pressure in the null region to poloidal magnetic pressure at the midplane. This transport also increases approximately linearly as d xx is reduced. A diffusive model of this transport is shown to predict the total transport across the null points, where diffusion coefficients of up to ∼ 10 2 m 2 s −1 centered on a small region around the X-points are used. However, the CM also results in significant changes to the flux surfaces in the null region, which is not captured by this diffusive model. The changes in magnetic geometry mean the fractional exhaust power delivered to each divertor leg is highly sensitive to β pm ⁠, d xx ⁠, and θ⁠. For small values of θ⁠, the CM can induce a change in topology, redirecting exhaust power from a secondary divertor leg on the high field side to one on the low field side. Similar behavior is found in the fraction of exhaust power going to the inner and outer divertor. Such changes in the flux surfaces may not be captured by Grad–Shafranov solvers and so may be a source of error in the magnetic reconstruction of SF experiments. We consistently find that the fractional exhaust power going to a secondary divertor leg on the high field side is small, consistent with SF experiments.

Computer simulation↗

A high-order Lagrangian discontinuous Galerkin hydrodynamic method for quadratic cells using a subcell mesh stabilization scheme

We present here a Lagrangian discontinuous Galerkin (DG) hydrodynamic method that is up to third-order accurate using subcell mesh stabilization (SMS) for compressible flows on quadratic meshes in two-dimensional (2D) Cartesian coordinates. Similar to the second-order accurate Lagrangian DG method with linear meshes, the physical evolution equations for the specific volume, velocity, and specific total energy are discretized using a modal DG method with Taylor series polynomials. The Riemann velocity at the vertices of a curvilinear cell, and the corresponding surface forces, are calculated by solving a multidirectional approximate Riemann problem. Curvilinear cells (e.g., quadratic quadrilateral meshes in this work) have many deformational degrees of freedom, and with these cells, they can deform in unphysical ways. Likewise, the Riemann solution at an edge vertex differs from the one at the corner of a cell. With SMS, each quadratic quadrilateral cell is decomposed into four quadrilateral subcells, that move in a Lagrangian manner. The edge vertex is surrounded by four subcells so that it is similar to the vertex at the cell corner. SMS can detect inconsistent density fields between the cell and subcells. The difference between these two density fields is used to correct the stress (pressure) input to the Riemann solver. This SMS scheme enables stable mesh motion and accurate solutions in the context of a Lagrangian high-order DG method that is up to third-order with quadratic cells. We also present effective limiting strategies that ensure monotonicity of the primitive variables with the high-order DG method. This Lagrangian DG hydrodynamic method with SMS conserves mass, momentum, and total energy. A suite of test problems are calculated to demonstrate the designed order of accuracy (up to third-order accurate) of this method, and that the Lagrangian DG method using SMS preserves cylindrical symmetry on 1D radial flow problems with an equal-angle polar quadratic mesh.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Software Synthesis for High Productivity Exascale Computing

Over the three years of our project, we accomplished three key milestones: We demonstrated how ideas from generative programming and software synthesis can help support the development of bulk-synchronous distributed memory kernels. These ideas are realized in a new language called MSL, a C-like language that combines synthesis features with high level notations for array manipulation and bulk-synchronous parallelism to simplify the semantic analysis required for synthesis. We also demonstrated that these high level notations map easily to low level C code and show that the performance of this generated code matches that of handwritten Fortran. Second, we introduced the idea of solver-aided domain-specific languages (SDSLs), which are an emerging class of computer-aided programming systems. SDSLs ease the construction of programs by automating tasks such as verification, debugging, synthesis, and non-deterministic execution. SDSLs are implemented by translating the DSL program into logical constraints. Next, we developed a symbolic virtual machine called Rosette, which simplifies the construction of such SDSLs and their compilers. We have used Rosette to build SynthCL, a subset of OpenCL that supports synthesis. Third, we developed novel numeric algorithms that move as little data as possible, either between levels of a memory hierarchy or between parallel processors over a network. We achieved progress in three aspects of this problem. First we determined lower bounds on communication. Second, we compared these lower bounds to widely used versions of these algorithms, and noted that these widely used algorithms usually communicate asymptotically more than is necessary. Third, we identified or invented new algorithms for most linear algebra problems that do attain these lower bounds, and demonstrated large speed-ups in theory and practice.

97 MATHEMATICS AND COMPUTING↗

Self-Excited, Multi-kHz Dynamics in a Linear, Semi-Bounded Detonation Channel

A linear, semi-bounded detonation channel was studied to characterize the injection, mixing, and ignition process in a two dimensional analog of a rotating detonation wave combustor (RDWC). The linear channel was developed to provide large optical accessibility in support of high-fidelity diagnostics. The canonical geometry was also chosen to support the generation of future numerical validation. A high degree of modularity was designed into the combustor to allow for testing multiple injector con figurations, and for reconfiguring the high-frequency instrumentation and optical window locations. Initial experiments have shown the presence of self-excited, multi-kHz combustion waves that bear the distinct characteristics of detonation waves. High-frequency pressure data details the dynamic response to detonation propagation in the combustor. The role of manifold-chamber coupling is investigated as the mechanism behind the formation of self-excited detonation waves in the combustor. We explore the formation of the waves using Schlieren imaging techniques and examine the structures present within the reaction zone and the interface between reactants and products. Companion computations were carried out using an in-house CFD solver GEMS(General Equations and Mesh Solver) to develop understanding of the structure and dynamics of the flow, flame, and their coupled interactions.

Schwinn, Kyle S.↗

An implementation analysis of the linear discontinuous finite element method

This paper provides an implementation analysis of the linear discontinuous finite element method (LD-FEM) that spans the space of (l, x, y, z). A practical implementation of LD includes 1) selecting a computationally efficient algorithm to solve the 4 x 4 matrix system Ax = b that describes the angular flux in a mesh element, and 2) choosing how to store the data used to construct the matrix A and the vector b to either reduce memory consumption or increase computational speed. To analyze the first of these, three algorithms were selected to solve the 4 x 4 matrix equation: Cramer's rule, a streamlined implementation of Gaussian elimination, and LAPACK's Gaussian elimination subroutine dgesv. The results indicate that Cramer's rule and the streamlined Gaussian elimination algorithm perform nearly equivalently and outperform LAPACK's implementation of Gaussian elimination by a factor of 2. To analyze the second implementation detail, three formulations of the discretized LD-FEM equations were provided for implementation in a transport solver: 1) a low-memory formulation, which relies heavily on 'on-the-fly' calculations and less on the storage of pre-computed data, 2) a high-memory formulation, which pre-computes much of the data used to construct A and b, and 3) a reduced-memory formulation, which lies between the low - and high-memory formulations. These three formulations were assessed in the Jaguar transport solver based on relative memory footprint and computational speed for increasing mesh size and quadrature order. The results indicated that the memory savings of the low-memory formulation were not sufficient to warrant its implementation. The high-memory formulation resulted in a significant speed advantage over the reduced-memory option (10-50%), but also resulted in a proportional increase in memory consumption (5-45%) for increasing quadrature order and mesh count; therefore, the practitioner should weigh the system memory constraints against any required runtime constraints to determine which option is more suitable for his or her specific applications. (authors)

97 MATHEMATICS AND COMPUTING↗

Efficient shallow Ritz method for 1D diffusion problems

This paper studies the shallow Ritz method for solving the one-dimensional diffusion problem. It is shown that the shallow Ritz method improves the order of approximation dramatically for non-smooth problems. To realize this optimal or nearly optimal order of the shallow Ritz approximation, we develop a damped block Newton (dBN) method that alternates between updates of the linear and non-linear parameters. Per each iteration, the linear and the non-linear parameters are updated by exact inversion and one step of a modified, damped Newton method applied to a reduced non-linear system, respectively. The computational cost of each dBN iteration is $\mathcal{O}$(n). Starting with the non-linear parameters as a uniform partition of the interval, numerical experiments show that the dBN is capable of efficiently moving mesh points to nearly optimal locations. In conclusion, to improve the efficiency of the dBN further, we propose an adaptive damped block Newton (AdBN) method by combining the dBN with the adaptive neuron enhancement (ANE) method [28].

Diffusion problems↗

Mixed-precision iterative refinement using tensor cores on GPUs to accelerate solution of linear systems

Double-precision floating-point arithmetic (FP64) has been the de facto standard for engineering and scientific simulations for several decades. Problem complexity and the sheer volume of data coming from various instruments and sensors motivate researchers to mix and match various approaches to optimize compute resources, including different levels of floating-point precision. In recent years, machine learning has motivated hardware support for half-precision floating-point arithmetic. A primary challenge in high-performance computing is to leverage reduced-precision and mixed-precision hardware. We show how the FP16/FP32 Tensor Cores on NVIDIA GPUs can be exploited to accelerate the solution of linear systems of equations Ax = b without sacrificing numerical stability. The techniques we employ include multiprecision LU factorization, the preconditioned generalized minimal residual algorithm (GMRES), and scaling and auto-adaptive rounding to avoid overflow. We also show how to efficiently handle systems with multiple right-hand sides. On the NVIDIA Quadro GV100 (Volta) GPU, we achieve a 4×-5× performance increase and 5× better energy efficiency versus the standard FP64 implementation while maintaining an FP64 level of numerical stability.

GMRES↗

Multi-Scale Initial Conditions For Cosmological Simulations

We discuss a new algorithm to generate multi-scale initial conditions with multiple levels of refinements for cosmological 'zoom-in' simulations. The method uses an adaptive convolution of Gaussian white noise with a real-space transfer function kernel together with an adaptive multi-grid Poisson solver to generate displacements and velocities following first- (1LPT) or second-order Lagrangian perturbation theory (2LPT). The new algorithm achieves rms relative errors of the order of 10{sup -4} for displacements and velocities in the refinement region and thus improves in terms of errors by about two orders of magnitude over previous approaches. In addition, errors are localized at coarse-fine boundaries and do not suffer from Fourier-space-induced interference ringing. An optional hybrid multi-grid and Fast Fourier Transform (FFT) based scheme is introduced which has identical Fourier-space behaviour as traditional approaches. Using a suite of re-simulations of a galaxy cluster halo our real-space-based approach is found to reproduce correlation functions, density profiles, key halo properties and subhalo abundances with per cent level accuracy. Finally, we generalize our approach for two-component baryon and dark-matter simulations and demonstrate that the power spectrum evolution is in excellent agreement with linear perturbation theory. For initial baryon density fields, it is suggested to use the local Lagrangian approximation in order to generate a density field for mesh-based codes that is consistent with the Lagrangian perturbation theory instead of the current practice of using the Eulerian linearly scaled densities.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗