Aether.jl : A High-Performance 3D MHD and Multifluid Dust Code Written in a Dynamic Language with an Interactive Human-AI Development Framework

Aether.jl is a new finite-volume code for compressible hydrodynamics and magnetohydrodynamics, written in Julia and primarily designed for GPU systems. It solves the MHD equations with constrained transport in Cartesian, cylindrical, and spherical-polar coordinates using standard high-order Godunov methods, and can couple an arbitrary number of dust fluids to the gas through stiff mutual drag.

Developed from scratch with an interactive human-AI coding-agent workflow, the paper documents this workflow alongside the numerical methods. Performance-critical kernels use KernelAbstractions.jl, enabling runs on CPUs and GPUs from multiple vendors; the code works either from interactive notebooks or batch scripts, keeping prototyping, production runs, and analysis in a single language.

The implementation is verified through a series of hydrodynamic, MHD, and dust tests. Although written in a dynamic language, Aether.jl achieves comparable or even higher single-GPU throughput than C++ code on the same hardware. In weak scaling on Frontier, parallel efficiency stays above 93% on 4096 GCDs. These results indicate that a dynamic language can now support production astrophysical MHD simulations on exascale systems; Aether.jl and its Jupyter notebook example suite are publicly available.