arXiv subjects

Lukas Einkemmer

Publications and source records attributed to Lukas Einkemmer.

At least 91 records · Page 5Linked to original sources

arXiv2017

A study on conserving invariants of the Vlasov equation in semi-Lagrangian computer simulations

The semi-Lagrangian discontinuous Galerkin method, coupled with a splitting approach in time, has recently been introduced for the Vlasov--Poisson equation. Since these methods are conservative, local in space, and able to limit numerical diffusion, they are considered a promising alternative to more traditional semi-Lagrangian schemes. In this paper we study the conservation of important physical invariants and the long time behavior of the semi-Lagrangian discontinuous Galerkin method. To that end we conduct a theoretical analysis and perform a number of numerical simulations. In particular, we find that the entropy is nondecreasing for the discontinuous Galerkin scheme, while unphysical oscillations in the entropy are observed for the traditional method based on cubic spline interpolation.

math.NA↗

arXiv2017

Streamline integration as a method for two-dimensional elliptic grid generation

We propose a new numerical algorithm to construct a structured numerical elliptic grid of a doubly connected domain. Our method is applicable to domains with boundaries defined by two contour lines of a two-dimensional function. The resulting grids are orthogonal to the boundary. Grid points as well as the elements of the Jacobian matrix can be computed efficiently and up to machine precision. In the simplest case we construct conformal grids, yet with the help of weight functions and monitor metrics we can control the distribution of cells across the domain. Our algorithm is parallelizable and easy to implement with elementary numerical methods. We assess the quality of grids by considering both the distribution of cell sizes and the accuracy of the solution to elliptic problems. Among the tested grids these key properties are best fulfilled by the grid constructed with the monitor metric approach.

math.NA↗

arXiv2016

Evaluation of the Partitioned Global Address Space (PGAS) model for an inviscid Euler solver

In this paper we evaluate the performance of Unified Parallel C (which implements the partitioned global address space programming model) using a numerical method that is widely used in fluid dynamics. In order to evaluate the incremental approach to parallelization (which is possible with UPC) and its performance characteristics, we implement different levels of optimization of the UPC code and compare it with an MPI parallelization on four different clusters of the Austrian HPC infrastructure (LEO3, LEO3E, VSC2, VSC3) and on an Intel Xeon Phi. We find that UPC is significantly easier to develop in compared to MPI and that the performance achieved is comparable to MPI in most situations. The obtained results show worse performance (on VSC2), competitive performance (on LEO3, LEO3E and VSC3), and superior performance (on the Intel Xeon Phi).

cs.DC↗

arXiv2016

ADI type preconditioners for the steady state inhomogeneous Vlasov equation

The purpose of the current work is to find numerical solutions of the steady state inhomogeneous Vlasov equation. This problem has a wide range of applications in the kinetic simulation of non-thermal plasmas. However, the direct application of either time stepping schemes or iterative methods (such as Krylov based methods like GMRES or relexation schemes) is computationally expensive. In the former case the slowest timescale in the system forces us to perform a long time integration while in the latter case a large number of iterations is required. In this paper we propose a preconditioner based on an ADI type splitting method. This preconditioner is then combined with both GMRES and Richardson iteration. The resulting numerical schemes scale almost ideally (i.e. the computational effort is proportional to the number of grid points). Numerical simulations conducted show that this can result in a speedup of close to two orders of magnitude (even for intermediate grid sizes) with respect to the not preconditioned case. In addition, we discuss the characteristics of these numerical methods and show the results for a number of numerical simulations.

physics.comp-ph↗

arXiv2016

An asymptotic preserving scheme for the relativistic Vlasov--Maxwell equations in the classical limit

We consider the relativistic Vlasov--Maxwell (RVM) equations in the limit when the light velocity $c$ goes to infinity. In this regime, the RVM system converges towards the Vlasov--Poisson system and the aim of this paper is to construct asymptotic preserving numerical schemes that are robust with respect to this limit. Our approach relies on a time splitting approach for the RVM system employing an implicit time integrator for Maxwell's equations in order to damp the higher and higher frequencies present in the numerical solution. A number of numerical simulations are conducted in order to investigate the performances of our numerical scheme both in the relativistic as well as in the classical limit regime. In addition, we derive the dispersion relation of the Weibel instability for the continuous and the discretized problem.

math.NA↗

arXiv2016

Structure preserving numerical methods for the Vlasov equation

To preserve a number of physically relevant invariants is a major concern when considering long time integration of the Vlasov equation. In the present work we consider the semi-Lagrangian discontinuous Galerkin method for the Vlasov-Poisson system. We discuss the performance of this method and compare it to cubic spline interpolation, where appropriate. In addition, numerical simulations for the two-stream instability are shown.

math.NA↗

arXiv2016

A mixed precision semi-Lagrangian algorithm and its performance on accelerators

In this paper we propose a mixed precision algorithm in the context of the semi-Lagrangian discontinuous Galerkin method. The performance of this approach is evaluated on a traditional dual socket workstation as well as on a Xeon Phi and an NVIDIA K80. We find that the mixed precision algorithm can be implemented efficiently on these architectures. This implies that, in addition to the considerable reduction in memory, a substantial increase in performance can be observed as well. Moreover, we discuss the relative performance of our implementations.

cs.MS↗

arXiv2016

Overcoming order reduction in diffusion-reaction splitting. Part 2: oblique boundary conditions

Splitting methods constitute a well-established class of numerical schemes for the time integration of partial differential equations. Their main advantages over more traditional schemes are computational efficiency and superior geometric properties. In the presence of non-trivial boundary conditions, however, splitting methods usually suffer from order reduction and some additional loss of accuracy. For diffusion-reaction equations with inhomogeneous oblique boundary conditions, a modification of the classic second-order Strang splitting is proposed that successfully resolves the problem of order reduction. The same correction also improves the accuracy of the classic first-order Lie splitting. The proposed modification only depends on the available boundary data and, in the case of non Dirichlet boundary conditions, on the computed numerical solution. Consequently, this modification can be implemented in an efficient way, which makes the modified splitting schemes superior to their classic versions. The framework employed in our error analysis also allows us to explain the fractional orders of convergence that are often encountered for classic Strang splitting. Numerical experiments that illustrate the theory are provided.

math.NA↗

arXiv2015

A splitting approach for the Kadomtsev--Petviashvili equation

We consider a splitting approach for the Kadomtsev--Petviashvili equation with periodic boundary conditions and show that the necessary interpolation procedure can be efficiently implemented. The error made by this numerical scheme is compared to exponential integrators which have been shown in Klein and Roidot (SIAM J. Sci. Comput., 2011) to perform best for stiff solutions of the Kadomtsev--Petviashvili equation. Since many classic high order splitting methods do not perform well, we propose a stable extrapolation method in order to construct an efficient numerical scheme of order four. In addition, the conservation properties and the possibility of order reduction for certain initial values for the numerical schemes under consideration is investigated.

physics.comp-ph↗

arXiv2015

High performance computing aspects of a dimension independent semi-Lagrangian discontinuous Galerkin code

The recently developed semi-Lagrangian discontinuous Galerkin approach is used to discretize hyperbolic partial differential equations (usually first order equations). Since these methods are conservative, local in space, and able to limit numerical diffusion, they are considered a promising alternative to more traditional semi-Lagrangian schemes (which are usually based on polynomial or spline interpolation). In this paper, we consider a parallel implementation of a semi-Lagrangian discontinuous Galerkin method for distributed memory systems (so-called clusters). Both strong and weak scaling studies are performed on the Vienna Scientific Cluster 2 (VSC-2). In the case of weak scaling, up to 8192 cores, we observe a parallel efficiency above 0.89 for both two and four dimensional problems. Strong scaling results show good scalability to at least 1024 cores (we consider problems that can be run on a single processor in reasonable time). In addition, we study the scaling of a two dimensional Vlasov--Poisson solver that is implemented using the framework provided. All of the simulation are conducted in the context of worst case communication overhead; i.e., in a setting where the CFL number increases linearly with the problem size. The framework introduced in this paper facilitates a dimension independent implementation (based on C++ templates) of scientific codes using both an MPI and a hybrid approach to parallelization. We describe the essential ingredients of our implementation.

physics.comp-ph↗

arXiv2014

Overcoming order reduction in diffusion-reaction splitting. Part 1: Dirichlet boundary conditions

For diffusion-reaction equations employing a splitting procedure is attractive as it reduces the computational demand and facilitates a parallel implementation. Moreover, it opens up the possibility to construct second-order integrators that preserve positivity. However, for boundary conditions that are neither periodic nor of homogeneous Dirichlet type order reduction limits its usefulness. In the situation described the Strang splitting procedure is not more accurate than Lie splitting. In this paper, we propose a splitting procedure that, while retaining all the favorable properties of the original method, does not suffer from order reduction. We demonstrate our results by conducting numerical simulations in one and two space dimensions with inhomogeneous and time dependent Dirichlet boundary conditions. In addition, a mathematical rigorous convergence analysis is conducted that confirms the results observed in the numerical simulations.

math.NA↗

arXiv2014

A modern resistive magnetohydrodynamics solver using C++ and the Boost library

In this paper we describe the implementation of our C++ resistive magnetohydrodynamics solver. The framework developed facilitates the separation of the code implementing the specific numerical method and the physical model, on the one hand, from the handling of boundary conditions and the management of the computational domain, on the other hand. In particular, this will allow us to use finite difference stencils which are only defined in the interior of the domain (the boundary conditions are handled automatically). We will discuss this and other design considerations and their impact on performance in some detail. In addition, we provide a documentation of the code developed and demonstrate that a performance comparable to Fortran can be achieved, while still maintaining a maximum of code readability and extensibility.

math.NA↗

arXiv2014

On the error propagation of semi-Lagrange and Fourier methods for advection problems

In this paper we study the error propagation of numerical schemes for the advection equation in the case where high precision is desired. The numerical methods considered are based on the fast Fourier transform, polynomial interpolation (semi-Lagrangian methods using a Lagrange or spline interpolation), and a discontinuous Galerkin semi-Lagrangian approach (which is conservative and has to store more than a single value per cell). We demonstrate, by carrying out numerical experiments, that the worst case error estimates given in the literature provide a good explanation for the error propagation of the interpolation-based semi-Lagrangian methods. For the discontinuous Galerkin semi-Lagrangian method, however, we find that the characteristic property of semi-Lagrangian error estimates (namely the fact that the error increases proportionally to the number of time steps) is not observed. We provide an explanation for this behavior and conduct numerical simulations that corroborate the different qualitative features of the error in the two respective types of semi-Lagrangian methods. The method based on the fast Fourier transform is exact but, due to round-off errors, susceptible to a linear increase of the error in the number of time steps. We show how to modify the Cooley--Tukey algorithm in order to obtain an error growth that is proportional to the square root of the number of time steps. Finally, we show, for a simple model, that our conclusions hold true if the advection solver is used as part of a splitting scheme.

math.NA↗

arXiv2014

A strategy to suppress recurrence in grid-based Vlasov solvers

In this paper we propose a strategy to suppress the recurrence effect present in grid-based Vlasov solvers. This method is formulated by introducing a cutoff frequency in Fourier space. Since this cutoff only has to be performed after a number of time steps, the scheme can be implemented efficiently and can relatively easily be incorporated into existing Vlasov solvers. Furthermore, the scheme proposed retains the advantage of grid-based methods in that high accuracy can be achieved. This is due to the fact that in contrast to the scheme proposed by Abbasi et al. no statistical noise is introduced into the simulation. We will illustrate the utility of the method proposed by performing a number of numerical simulations, including the plasma echo phenomenon, using a discontinuous Galerkin approximation in space and a Strang splitting based time integration.

physics.comp-ph↗

arXiv2014

A Hamiltonian splitting for the Vlasov-Maxwell system

A new splitting is proposed for solving the Vlasov-Maxwell system. This splitting is based on a decomposition of the Hamiltonian of the Vlasov-Maxwell system and allows for the construction of arbitrary high order methods by composition (independent of the specific deterministic method used for the discretization of the phase space). Moreover, we show that for a spectral method in space this scheme satisfies Poisson's equation without explicitly solving it. Finally, we present some examples in the context of the time evolution of an electromagnetic plasma instability which emphasizes the excellent behavior of the new splitting compared to methods from the literature.

math.NA↗

arXiv2013

An almost symmetric Strang splitting scheme for the construction of high order composition methods

In this paper we consider splitting methods for nonlinear ordinary differential equations in which one of the (partial) flows that results from the splitting procedure can not be computed exactly. Instead, we insert a well-chosen state $y_{\star}$ into the corresponding nonlinearity $b(y)y$, which results in a linear term $b(y_{\star})y$ whose exact flow can be determined efficiently. Therefore, in the spirit of splitting methods, it is still possible for the numerical simulation to satisfy certain properties of the exact flow. However, Strang splitting is no longer symmetric (even though it is still a second order method) and thus high order composition methods are not easily attainable. We will show that an iterated Strang splitting scheme can be constructed which yields a method that is symmetric up to a given order. This method can then be used to attain high order composition schemes. We will illustrate our theoretical results, up to order six, by conducting numerical experiments for a charged particle in an inhomogeneous electric field, a post-Newtonian computation in celestial mechanics, and a nonlinear population model and show that the methods constructed yield superior efficiency as compared to Strang splitting. For the first example we also perform a comparison with the standard fourth order Runge--Kutta methods and find significant gains in efficiency as well better conservation properties.

math.NA↗

arXiv2013

A conservative discontinuous Galerkin scheme for the 2D incompressible Navier--Stokes equations

In this paper we consider a conservative discretization of the two-dimensional incompressible Navier--Stokes equations. We propose an extension of Arakawa's classical finite difference scheme for fluid flow in the vorticity-stream function formulation to a high order discontinuous Galerkin approximation. In addition, we show numerical simulations that demonstrate the accuracy of the scheme and verify the conservation properties, which are essential for long time integration. Furthermore, we discuss the massively parallel implementation on graphic processing units.

physics.comp-ph↗

arXiv2013

Exponential Integrators on Graphic Processing Units

In this paper we revisit stencil methods on GPUs in the context of exponential integrators. We further discuss boundary conditions, in the same context, and show that simple boundary conditions (for example, homogeneous Dirichlet or homogeneous Neumann boundary conditions) do not affect the performance if implemented directly into the CUDA kernel. In addition, we show that stencil methods with position-dependent coefficients can be implemented efficiently as well. As an application, we discuss the implementation of exponential integrators for different classes of problems in a single and multi GPU setup (up to 4 GPUs). We further show that for stencil based methods such parallelization can be done very efficiently, while for some unstructured matrices the parallelization to multiple GPUs is severely limited by the throughput of the PCIe bus.

math.NA↗