AsterX: a new open-source GPU-accelerated GRMHD code for dynamical spacetimes
ORAL
Abstract
As the demand for extensive parallel computing grows to tackle various astrophysical challenges through numerical simulations, the necessity for codes capable of efficient GPU utilization becomes increasingly apparent. In this presentation, I will delve into the distinctive features of AsterX, a novel open-source code for general relativistic magnetohydrodynamics (GRMHD). AsterX is built upon CarpetX, a cutting-edge driver within the Einstein Toolkit. It employs the flux-conservative Valencia formulation, incorporating staggered vector potential evolution. Tailored for GPU performance, AsterX maximizes the advantages of block-structured adaptive mesh refinement, seamlessly integrated via CarpetX and the AMReX framework. While still in development, AsterX undergoes rigorous testing on high-performance clusters like Frontera and Frontier, demonstrating successful completion of some demanding GRMHD tests.
*We acknowledges support from NSF awards NSF AST-2009330, OAC-2031744 and PHY-1806596, PHY-2110352, NASA TCAN awards (NNH17ZDA001N and 80NSSC24K0100).Computational resources were provided by the TACC's Frontera supercomputer allocation No. PHY-20010 and AST-20021.Additional resources were provided by the RIT's BlueSky and Green Pairie and Lagoon Clusters acquired with NSF grants PHY-2018420, PHY-0722703, PHY-1229173 and PHY-1726215.
–
Presenters
-
Liwei Ji
- Rochester Institute of Technology