IllinoisGRMHD: An Open-Source, User-Friendly GRMHD Code for Dynamical Spacetimes

TL;DR

IllinoisGRMHD is an open-source, user-friendly, and efficient GRMHD code that facilitates community involvement in modeling dynamical spacetimes involving black holes and neutron stars.

Contribution

It is a highly-extensible, open-source rewrite of a previous closed-source code, reducing the learning curve and improving computational efficiency.

Findings

Nearly twice as fast as the original code at large scales

Produces identical output to the original code on adaptive-mesh grids

Facilitates community involvement and further development

Abstract

In the extreme violence of merger and mass accretion, compact objects like black holes and neutron stars are thought to launch some of the most luminous outbursts of electromagnetic and gravitational wave energy in the Universe. Modeling these systems realistically is a central problem in theoretical astrophysics, but has proven extremely challenging, requiring the development of numerical relativity codes that solve Einstein's equations for the spacetime, coupled to the equations of general relativistic (ideal) magnetohydrodynamics (GRMHD) for the magnetized fluids. Over the past decade, the Illinois Numerical Relativity (ILNR) Group's dynamical spacetime GRMHD code has proven itself as a robust and reliable tool for theoretical modeling of such GRMHD phenomena. However, the code was written "by experts and for experts" of the code, with a steep learning curve that would severely hinder community adoption if it were open-sourced. Here we present IllinoisGRMHD, which is an open-source, highly-extensible rewrite of the original closed-source GRMHD code of the ILNR Group. Reducing the learning curve was the primary focus of this rewrite, with the goal of facilitating community involvement in the code's use and development, as well as the minimization of human effort in generating new science. IllinoisGRMHD also saves computer time, generating roundoff-precision identical output to the original code on adaptive-mesh grids, but nearly twice as fast at scales of hundreds to thousands of cores.