Portal de Periódicos da CAPES

Simulation of Supernova Explosion Accelerated on GPU: Spherically Symmetric Neutrino-Radiation Hydrodynamics

2018; Springer Science+Business Media; Linguagem: Inglês

10.1007/978-3-319-95168-3_30

1611-3349

Hideo Matsufuru, Kohsuke Sumiyoshi,

Gamma-ray bursts and supernovae

Performing large scale numerical simulations is essential to understand the explosion mechanism of core-collapse supernovae. It is mandatory to solve a multi-physics system described by coupled equations of hydrodynamics and neutrino-radiation transfer in multi-dimensions. Since the neutrino transfer is in principle governed by the Boltzmann equation in six-dimensional space, numerical simulations require large computational resources. Having the final goal to realize acceleration of such simulations, we study the numerical computations of neutrino-radiation hydrodynamics under spherical symmetry by exploiting GPU devices, which has become a powerful equipment in scientific high performance computing. As the first step, we focus on the most time consuming part: a linear equation solver for a block tridiagonal matrix which appears in a spherically symmetric simulation. To offload this part to the GPU devices, we employ OpenACC as a framework of implementation as well as make use of cuBLAS library that is available on NVIDIA’s CUDA environment. Practical performance is examined on two systems with the Kepler and Pascal generations of NVIDIA’s GPU architecture.