Gpupegas: A New Gpu-Accelerated Hydrodynamic Code for Numerical Simulations of Interacting Galaxies

Abstract

In this paper a new scalable hydrodynamic code GPUPEGAS (GPUaccelerated PErformance Gas Astrophysic Simulation) for simulation of interacting galaxies is proposed. The code is based on combination of Godunov method as well as on the original implementation of FlIC method, specially adapted for GPU-implementation. Fast Fourier Transform is used for Poisson equation solution in GPUPEGAS. Software implementation of the above methods was tested on classical gas dynamics problems, new Aksenov's test and classical gravitational gas dynamics problems. Collisionless hydrodynamic approach was used for modelling of stars and dark matter. The scalability of GPUPEGAS computational accelerators is shown.

Co-design of Parallel Numerical Methods for Plasma Physics and Astrophysics

“…9 showing the simlation of the collision of two galaxies according to the described two-phase model. The details of the problem statement and simulation results can be found in [35]. …”

“…This approach has some limitations if it is necessary to trace a single particle and not a group of particles (for example, in the study of planet formation). But for the simulation of interacting galaxies this approach was successfully tested [35]. As it was already said in the introduction, the statement of the problem for the galactic objects dynamics is the solution of the equation of the gravitational magnetic gas dynamics for the gas component considering magnetic field and self-gravity and also the solution of the equations for the first moments of the collisionless Boltzmann equations.…”

“…These methods combine the advantages of both Lagrangian and Eulerian approaches. For a number of years the authors have been developing the Lagrangian-Eulerian approach on the basis of the combination of the Fluid-In-Cells method and Godunov method [9,11,34,35]. The system of gas dynamics equations is solved in two stages.…”

“…Such an approach makes it possible to use the same numerical algorithm [34] for the solution of magnetic gas dynamics equations, and also for the solution of the first moments of the Boltzmann equation [35]. This approach was efficiently implemented for two types of hybrid supercomputers: for the supercomputer with GPUs [35], and the supercomputer with Intel Xeon Phi accelerators [11].…”

Multilevel Parallelization: Grid Methods for Solving Direct and Inverse Problems

“…A comparison of different codes for subsonic turbulence is presented in [13]. Classical methods for simulation of magnetohydrodynamic turbulence such as adaptive mesh refinement (AMR) and smoothed-particle hydrodynamics are still widely used, but in recent years an impressive range of new methods have been proposed (a good review of these methods can be found in [14][15][16]). …”