Massively Parallel Computation of Accurate Densities for N-body Dark Matter Simulations using the Phase-Space-Element Method
Kavli Institute for Particle Astrophysics and Cosmology, SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA
arXiv:1612.09491 [physics.comp-ph], (29 Dec 2016)
@article{kaehler2016massively,
title={Massively Parallel Computation of Accurate Densities for N-body Dark Matter Simulations using the Phase-Space-Element Method},
author={Kaehler, Ralf},
year={2016},
month={dec},
archivePrefix={"arXiv"},
primaryClass={physics.comp-ph}
}
In 2012 a method to analyze N-body dark matter simulations using a tetrahedral tesselation of the three-dimensional dark matter manifold in six-dimensional phase space was introduced. This paper presents an accurate density computation approach for large N-body datasets, that is based on this technique and designed for massively parallel GPU-clusters. The densities are obtained by intersecting the tessellation with the cells of a spatially adaptive grid structure. We speed up this computational expensive part with an intersection algorithm, that is tailored to modern GPU architectures. We discuss different communication and dynamic load-balancing strategies and compare their weak and strong scaling efficiencies for several large N-body simulations.
January 4, 2017 by hgpu
