DM4 - Self Gravitating Systems and Dark Matter |
|
Speaker |
Spurzem, Rainer |
|
Coauthors |
Berczik, Peter; Silk Road Project Team |
| Talk Title |
Supermassive Black Hole Binaries in Galactic Nuclei and Gravitational Wave Emission simulated on large GPU Clusters |
|
Abstract |
Computer simulations of dense star clusters with and without central black holes done by team members of the Silk Road Project at NAOC (http://silkroad.bao.ac.cn) in Beijing are presented. They are applied to globular clusters and galactic nuclei. Many if not all galaxies contain supermassive black holes (SMBH). We study their growth and evolution by tidal accretion of stars and their dynamical interaction with the stellar environment. Also preliminary results from large direct N-body models of globular clusters will be shown. Relativistic corrections to Newtonian dynamics (so-called Post-Newtonian approximations) have to be taken into account to follow the dynamics of binaries with compact remnants (black holes, neutron stars). Spin-Spin and Spin-Orbit interactions are also considered. We discuss the gravitational wave emission of stellar mass black holes and supermassive black holes through ground based or space based current and future detectors. Gravitational wave astronomy will become in the future as rich as electromagnetic astronomy. The use of accelerated computing (currently: GPU) for astrophysical computations at National Astronomical Observatories of Chinese Academy of Sciences is presented. Current and future hardware (including a few preliminary experiments with Intel Xeon Phi) and international collaborations are discussed. We discuss many core accelerated particle- and mesh-based simulation algorithms for astrophysics, some of them using hundreds to thousands of GPUs (approaching a million GPU cores) for one single application run in a parallel message passing environment, and give detailed timing models. Future perspectives for GPU and FPGA accelerated computing towards Exaflop/s will be discussed for the field of astrophysical many-body simulations with gravity. In some more detail our two main parallel GPU N-body codes using high order Hermite schemes (NBODY6++/GPU and phiGPU) will be presented. |
|
Talk view |
|
|
|
|