Presentation
(RP08) News from Black Holes in Star Clusters Simulated with Large GPU Clusters
SessionResearch Posters Session
Event Type
Research Poster
Computer Architecture
MPI
Parallel Algorithms
Scientific Software Development
TimeTuesday, June 18th8:30am - 10am
LocationSubstanz 1, 2
DescriptionWe use the newly developed direct N-body integrator Nbody6++GPU to simulate the evolution of globular clusters with a high level of realism. We follow the dynamical and stellar evolution of individual and binary stars, their evolution into neutron stars and black holes, interactions, dynamical binary formation, and mergers. To accomplish this, Nbody6++GPU is parallelised using hybrid methods (MPI, GPU, OpenMP, and AVX/SSE) and reaches high performance for accurate gravity computation.
Nbody6++GPU has been used to compute the largest GC simulation so far, with one million stars. We present scaling and performance properties of the gravity calculation on multi-GPU systems and different single GPUs: Kepler K20, Pascal P100, and
Volta V100. In a scientific application, we study the formation of intermediate-mass black holes in globular star clusters, growing to more than a few hundred solar masses by interactions and mergers with other stars and stellar black holes. Typical formation paths are presented for simulations carried out at the newly installed COBRA GPU system of the Max Planck Computing & Data Facility in Garching.
Nbody6++GPU has been used to compute the largest GC simulation so far, with one million stars. We present scaling and performance properties of the gravity calculation on multi-GPU systems and different single GPUs: Kepler K20, Pascal P100, and
Volta V100. In a scientific application, we study the formation of intermediate-mass black holes in globular star clusters, growing to more than a few hundred solar masses by interactions and mergers with other stars and stellar black holes. Typical formation paths are presented for simulations carried out at the newly installed COBRA GPU system of the Max Planck Computing & Data Facility in Garching.
Poster PDF
Poster Author
