GRaM-X: A new GPU-accelerated dynamical spacetime GRMHD code for Exascale computing with the Einstein Toolkit

Abstract We present GRaM-X ( G eneral R elativistic a ccelerated M agnetohydrodynamics on AMRe X ), new GPU-accelerated dynamical-spacetime general relativistic magnetohydrodynamics (GRMHD) code which extends the GRMHD capability of Einstein Toolkit to GPU-based exascale systems. supports 3D adaptive mesh refinement (AMR) GPUs via AMR driver for called CarpetX in turn leverages AMReX , an library developed use by United States DOE’s Exascale Computing Project. Z4c formalism evolve equations and Valencia formulation GRMHD. both analytic as well tabulated state. implement TVD WENO reconstruction methods HLLE Riemann solver. test accuracy using range tests static spacetime, e.g. 1D shocktubes, 2D magnetic rotor cylindrical explosion, dynamical spacetimes, i.e. oscillations Tolman-Oppenheimer-Volkhof star. find excellent agreement with results other codes reported literature. also perform scaling that shows weak efficiency ∼40%–50% 2304 nodes (13824 NVIDIA V100 GPUs) respect single-node performance OLCF’s supercomputer Summit.