Large Scale Nuclear Pasta Calculations

Abstract:  In nuclear astro-physics, the quantum simulation of large inhomogeneous dense systems as present in the crusts of neutron stars presents big challenges. The feasible number of particles in a simulation box with periodic boundary conditions is strongly limited due to the immense computational cost of the quantum methods. In this talk, we describe the techniques used to parallelize a state of the art density functional theory code that operates on an equidistant grid, and optimize its performance on distributed memory architectures. We also describe techniques to accelerate the compute-intensive matrix calculation part in this formalism. Presented techniques allow us to achieve good scaling and high performance on a large number of cores, as demonstrated through detailed performance analysis on Edison, a Cray XC30 supercomputer. Furthermore we show first applications simulating so called pasta phases with up to 6000 particles which was not possible without a highly parallelized code.