GRHayL: An Infrastructure-Agnostic GRMHD Library for Reproducible Multimessenger Numerical Relativity

Progress in multimessenger astrophysics hinges on reproducible numerical relativity, which in turn needs cross-code validation and rapid method sharing. GRHayL is an open-source, infrastructure-agnostic GRMHD library that exposes clean C/C++ APIs and composable kernels. It implements conservative-to-primitive recovery with a multi-stage backup ladder, PPM/PLM reconstruction with an HLL Riemann solver for accurate shock capturing, vector-potential formulation to maintain divergence-free magnetic fields and ensure long-term stability, and fourth-order geometric source terms for curved-spacetime balance. The same kernels run unchanged in Einstein Toolkit/Carpet, CarpetX (AMReX-based), and BlackHoles@Home/NRPy. In Carpet, refactoring reduces GRMHD grid function counts by ~40%, yielding substantially smaller memory footprints, and fourth-order sources halve central-density drift in TOV evolutions. Magnetized BNS simulations show excellent consistency with IllinoisGRMHD, with differences matching those induced by roundoff-level perturbations of the initial data. GRHayL is CPU-optimized today and structured for device portability, with CarpetX offload in progress. We will present ongoing work on higher-order finite-volume schemes, tabulated equations of state, and M1 neutrino/radiation transport, advancing a shared foundation for more realistic, scalable multimessenger simulations across diverse infrastructures.