Develop a Julia-native GPU-first astrophysical MHD code ecosystem

Develop a Julia-native, GPU-first finite-volume code for compressible astrophysical magnetohydrodynamics, extending high-performance production MHD simulation capabilities within a dynamic-language environment.

Background

The paper identifies the development of a Julia-native, GPU-first finite-volume solver for compressible astrophysical magnetohydrodynamics as an unresolved direction. Existing production astrophysical codes are predominantly written in statically compiled languages, while Julia offers a dynamic, interactive programming model combined with just-in-time compilation and support for heterogeneous hardware.

Aether.jl addresses this direction by providing such a solver and demonstrating its performance and scalability. However, the authors present the broader development of this class of software as an open direction rather than as a fully settled research problem, particularly with respect to the maturation of a complete production ecosystem for astrophysical MHD in Julia.