CUBE: An Informationoptimized Parallel Cosmological Nbody Algorithm
Abstract
Cosmological largescale structure Nbody simulations are computationlight, memoryheavy problems in supercomputing. The considerable amount of memory is usually dominated by an inefficient way of storing more than sufficient phase space information of particles. We present a new parallel, informationoptimized, particlemeshbased Nbody code CUBE, in which informationefficiency and memoryefficiency are increased by nearly an order of magnitude. This is accomplished by storing particle’s relative phase space coordinates instead of global values, and in the format of fixed point as light as 1 byte. The remaining information is given by complementary density and velocity fields (negligible in memory space) and proper ordering of particles (no extra memory). Our numerical experiments show that this informationoptimized Nbody algorithm provides accurate results within the error of the particlemesh algorithm. This significant lowering of the memorytocomputation ratio breaks the bottleneck of scaling up and speeding up large cosmological Nbody simulations on multicore and heterogeneous computing systems.
 Publication:

The Astrophysical Journal Supplement Series
 Pub Date:
 August 2018
 DOI:
 10.3847/15384365/aac830
 arXiv:
 arXiv:1712.06121
 Bibcode:
 2018ApJS..237...24Y
 Keywords:

 largescale structure of universe;
 methods: numerical;
 Astrophysics  Cosmology and Nongalactic Astrophysics;
 Astrophysics  Instrumentation and Methods for Astrophysics
 EPrint:
 10 pages, 9 figures, 2 tables. Accepted by ApJS