-
LeXInt: GPU-accelerated Exponential Integrators package
Authors:
Pranab J Deka,
Alexander Moriggl,
Lukas Einkemmer
Abstract:
We present an open-source CUDA-based package that consists of a compilation of exponential integrators where the action of the matrix exponential or the $\varphi_l$ functions on a vector is approximated using the method of polynomial interpolation at Leja points. Using a couple of test examples on an NVIDIA A100 GPU, we show that one can achieve significant speedups using CUDA over the correspondi…
▽ More
We present an open-source CUDA-based package that consists of a compilation of exponential integrators where the action of the matrix exponential or the $\varphi_l$ functions on a vector is approximated using the method of polynomial interpolation at Leja points. Using a couple of test examples on an NVIDIA A100 GPU, we show that one can achieve significant speedups using CUDA over the corresponding CPU code. LeXInt, written in a modular format, facilitates easy integration into any existing software package, and can be used for temporal integration of any differential equation.
△ Less
Submitted 22 October, 2024; v1 submitted 12 October, 2023;
originally announced October 2023.
-
Efficient numerical methods for Anisotropic Diffusion of Galactic Cosmic Rays
Authors:
Pranab J. Deka,
Ralf Kissmann,
Lukas Einkemmer
Abstract:
Anisotropic diffusion is imperative in understanding cosmic ray diffusion across the Galaxy, the heliosphere, and the interplay of cosmic rays with the Galactic magnetic field. This diffusion term contributes to the highly stiff nature of the cosmic ray transport equation. To conduct numerical simulations of time-dependent cosmic ray transport, implicit integrators (namely, Crank-Nicolson (CN)) ha…
▽ More
Anisotropic diffusion is imperative in understanding cosmic ray diffusion across the Galaxy, the heliosphere, and the interplay of cosmic rays with the Galactic magnetic field. This diffusion term contributes to the highly stiff nature of the cosmic ray transport equation. To conduct numerical simulations of time-dependent cosmic ray transport, implicit integrators (namely, Crank-Nicolson (CN)) have been traditionally favoured over the CFL-bound explicit integrators in order to be able to take large step sizes. We propose exponential methods to treat the linear anisotropc diffusion equation in the presence of advection and time-independent and time-dependent sources. These methods allow us to take even larger step sizes that can substantially speed-up the simulations whilst generating highly accurate solutions. In or subsequent work, we will use these exponential solvers in the Picard code to study anisotropic cosmic ray diffusion and we will consider additional physical processes such as continuous momentum losses and reacceleration.
△ Less
Submitted 8 August, 2023; v1 submitted 23 July, 2023;
originally announced July 2023.
-
Exponential methods for anisotropic diffusion
Authors:
Pranab J. Deka,
Lukas Einkemmer,
Ralf Kissmann
Abstract:
The anisotropic diffusion equation is imperative in understanding cosmic ray diffusion across the Galaxy, the heliosphere, and its interplay with the ambient magnetic field. This diffusion term contributes to the highly stiff nature of the CR transport equation. In order to conduct numerical simulations of time-dependent cosmic ray transport, implicit integrators have been traditionally favoured o…
▽ More
The anisotropic diffusion equation is imperative in understanding cosmic ray diffusion across the Galaxy, the heliosphere, and its interplay with the ambient magnetic field. This diffusion term contributes to the highly stiff nature of the CR transport equation. In order to conduct numerical simulations of time-dependent cosmic ray transport, implicit integrators have been traditionally favoured over the CFL-bound explicit integrators in order to be able to take large step sizes. We propose exponential methods that directly compute the exponential of the matrix to solve the linear anisotropic diffusion equation. These methods allow us to take even larger step sizes; in certain cases, we are able to choose a step size as large as the simulation time, i.e., only one time step. This can substantially speed-up the simulations whilst generating highly accurate solutions (l2 error $\leq 10^{-10}$). Additionally, we test an approach based on extracting a constant diffusion coefficient from the anisotropic diffusion equation, where the constant coefficient term is solved implicitly or exponentially and the remainder is treated using some explicit method. We find that this approach, for homogeneous linear problems, is unable to improve on the exponential-based methods that directly evaluate the matrix exponential.
△ Less
Submitted 12 December, 2022; v1 submitted 16 November, 2022;
originally announced November 2022.
-
A comparison of Leja- and Krylov-based iterative schemes for Exponential Integrators
Authors:
Pranab J. Deka,
Mayya Tokman,
Lukas Einkemmer
Abstract:
Krylov-based algorithms have long been preferred to compute the matrix exponential and exponential-like functions appearing in exponential integrators. Of late, direct polynomial interpolation of the action of these exponential-like functions have been shown to be competitive with the Krylov methods. We analyse the performance of the state-of-the-art Krylov algorithm, KIOPS, and the method of poly…
▽ More
Krylov-based algorithms have long been preferred to compute the matrix exponential and exponential-like functions appearing in exponential integrators. Of late, direct polynomial interpolation of the action of these exponential-like functions have been shown to be competitive with the Krylov methods. We analyse the performance of the state-of-the-art Krylov algorithm, KIOPS, and the method of polynomial interpolation at Leja points for a number of exponential integrators for various test problems and with varying amounts of stiffness. Additionally, we investigate the performance of an iterative scheme that combines both the KIOPS and Leja approach, named LeKry, that shows substantial improvements over both the Leja- and Krylov-based methods for certain exponential integrators. Whilst we do manage to single out a favoured iterative scheme for each of the exponential integrators that we consider in this study, we do not find any conclusive evidence for preferring either KIOPS or Leja for different classes of exponential integrators. We are unable to identify a superior exponential integrator, one that performs better than all others, for most, if not all of the problems under consideration. We, however, do find that the performance significantly depends on the interplay between the iterative scheme and the specific exponential integrator under consideration.
△ Less
Submitted 16 November, 2022;
originally announced November 2022.
-
LeXInt: Package for Exponential Integrators employing Leja interpolation
Authors:
Pranab J. Deka,
Lukas Einkemmer,
Mayya Tokman
Abstract:
We present a publicly available software for exponential integrators that computes the $\varphi_l(z)$ functions using polynomial interpolation. The interpolation method at Leja points have recently been shown to be competitive with the traditionally-used Krylov subspace method. The developed framework facilitates easy adaptation into any Python software package for time integration.
We present a publicly available software for exponential integrators that computes the $\varphi_l(z)$ functions using polynomial interpolation. The interpolation method at Leja points have recently been shown to be competitive with the traditionally-used Krylov subspace method. The developed framework facilitates easy adaptation into any Python software package for time integration.
△ Less
Submitted 20 January, 2023; v1 submitted 17 August, 2022;
originally announced August 2022.
-
Efficient adaptive step size control for exponential integrators
Authors:
Pranab Jyoti Deka,
Lukas Einkemmer
Abstract:
Traditional step size controllers make the tacit assumption that the cost of a time step is independent of the step size. This is reasonable for explicit and implicit integrators that use direct solvers. In the context of exponential integrators, however, an iterative approach, such as Krylov methods or polynomial interpolation, to compute the action of the required matrix functions is usually emp…
▽ More
Traditional step size controllers make the tacit assumption that the cost of a time step is independent of the step size. This is reasonable for explicit and implicit integrators that use direct solvers. In the context of exponential integrators, however, an iterative approach, such as Krylov methods or polynomial interpolation, to compute the action of the required matrix functions is usually employed. In this case, the assumption of constant cost is not valid. This is, in particular, a problem for higher-order exponential integrators, which are able to take relatively large time steps based on accuracy considerations. In this paper, we consider an adaptive step size controller for exponential Rosenbrock methods that determines the step size based on the premise of minimizing computational cost. The largest allowed step size, given by accuracy considerations, merely acts as a constraint. We test this approach on a range of nonlinear partial differential equations. Our results show significant improvements (up to a factor of 4 reduction in the computational cost) over the traditional step size controller for a wide range of tolerances.
△ Less
Submitted 17 August, 2022; v1 submitted 4 February, 2021;
originally announced February 2021.
