HyperCloud.jl

HyperCloud.jl is a meshfree framework for hyperbolic conservation laws. It solves PDEs on particle clouds with moving-least-squares (MLS) reconstructions: first-order upwind, MUSCL up to fifth order, and WENO. Slope limiters and a posteriori MOOD limiting keep shocks stable. A kinetic relaxation pipeline with IMEX time stepping approximates nonlinear systems by linear transport plus a stiff source term.

HyperCloud is part of the PDEStudio ecosystem:

• HyperCloudCore.jl provides the numerical engine (particle grids, MLS schemes, MOOD, time steppers, boundary conditions, GPU backends) and the config-driven builders that HyperCloud's runners call. HyperCloud re-exports it.
• PDEStudioCore.jl runs batches of simulations, caches results on disk under a hash of their parameters, and computes error statistics.
• PDEStudio.jl plots the results, either locally (GLMakie) or in the browser from a headless machine (WGLMakie + Bonito).

A first simulation

A simulation is a function returning a SimulationConfig: shared parameters, a dictionary of numerical methods, and the methods to run.

using HyperCloud

function my_advection()
    shared = Dict{Symbol, Any}(
        :sim_func_name => :run_direct_simulation,
        :tmax => 1.0, :snapshots => 20,

        :PDE_name => :linear, :PDE_velocities => ((1.0,),),   # u_t + u_x = 0

        :Grid_domain => :rectangular, :Grid_mins => (0.0,), :Grid_maxs => (1.0,),
        :Grid_Ns => (200,), :Grid_periodic => true,
        :Grid_randomness_factor => (0.2,), :Grid_seed => 42, :Grid_mover => :none,

        :Weight_name => :exponential, :Weight_alpha => 1.0, :Weight_range => 5.5,

        :IC_name => :gauss, :IC_a => (1.0,), :IC_b => (0.5,), :IC_width => 0.1,

        :Time_CFL => 0.2,
    )

    methods = Dict{Symbol, Dict{Symbol, Any}}(
        :analytical_solution => Dict{Symbol, Any}(),
        :RK2MUSCL2 => last(create_method(:RK2, :MUSCL, 2)),
        :RK4MUSCL4 => last(create_method(:RK4, :MUSCL, 4)),
    )

    return SimulationConfig(shared, methods, [:analytical_solution, :RK2MUSCL2, :RK4MUSCL4];
                            ref_func_name = :analytical_solution)
end

@sim my_advection     # run and cache
@plot my_advection    # open the interactive viewer (needs PDEStudio and a Makie backend)

Every available key is listed in the Configuration Reference.

Workflow

1. Configure. Write a function returning a SimulationConfig, either in your script or as name.jl in one of the CONFIG_SEARCH_PATHS folders (see Execution Macros).
2. Run. @sim name builds and runs every active method and caches the results.
3. Visualize. @plot name loads the configuration into the PDEStudio GUI.
4. Archive. @backup name copies the configuration file into a timestamped backups folder.

Package layout