This article is cited in 2 papers

Supercomputer modeling

Numerical hydrodynamics simulation of astrophysical flows at Intel Xeon Phi supercomputers

I. M. Kulikov^a, I. G. Chernykh^a, E. I. Vorobyov^b, A. V. Snytnikov^a, D. V. Weins^a, A. A. Moskovsky^c, A. B. Shmelev^c, V. A. Protasov^d, A. A. Serenko^d, V. E. Nenashev^d, V. A. Vshivkov^a, A. S. Rodionov^a, B. M. Glinsky^a, A. V. Tutukov<sup>e

a</sup> Institute of Computational Mathematics and Mathematical Geophysics SB RAS (pr. Ac. Lavryenteva 6, Novosibirsk, 630090 Russia)
^b Southern Federal University (pr. Stachki 194, Rostov-on-Don, 344090 Russia)
^c ZAO RSC Technologies (Kutuzovskiy pr. 36, building 23, Moscow, 121170 Russia)
^d Novosibirsk State Technical University (pr. K. Marksa 20, Novosibirsk, 630073 Russia)
^e Institute of Astronomy RAS (Pyatnitskaya St. 48, Moscow, 119017 Russia)

Abstract: In this paper we propose a research of AstroPhi code for numerical simulation of astrophysical flows at Intel Xeon Phi supercomputers. The co-design of a computational astrophysics model are described. The parallel implementation and scalability tests of the AstroPhi code are presented. The results of simulation of interaction between intergalactic wind and a disk galaxy are provided. For AstroPhi code a $134$x speed-up with one Intel Xeon Phi accelerator and $75 \%$ weak scaling efficiency on $224$x Intel Xeon Phi accelerators was obtained. We got peak of performance on a $7168 \times 1024 \times 1024$ mesh size by means $53760$ RSC PetaStream threads.

Keywords: high performance computing, numerical astrophysics, Intel Xeon Phi accelerators.

UDC: 519.63, 524.3

Received: 13.04.2016

DOI: 10.14529/cmse160406

Bibliographic databases:

•