Diffusion Limited Aggregation in Rust

Just to play with something while testing the ɴsɪ crate.

2k resolution, 66,666 instances of a scaffoldy dodecahedron. 512 shading, 64 aa-samples. Rendered in about one minute via 3Delight Cloud. Post-processing in Darktable.

Dependencies

This crate depends on nsi-sys which in term requires a renderer that implements the ɴsɪ API. Currently the only renderer that does is 3Delight which, though commercial, has been and is free for personal use since over twenty years.

Prerequisites

Before you start, download a 3Delight package for your platform & install it (supported: Linux, macOS, Windows). This will set the $DELIGHT environment variable that the build script is looking for to find headers and the library to link against.

Note: The free version of 3Delight will render with up to 12 cores on your machine. For crazier projects you can use their cheap cloud rendering service that gives you access to unlimited CPU cores. When you register you get 1,000 cloud minutes for free which ain’t too shabby and may come in handy once you start experimenting with this codebase yourself.

Building

The space partitioning insertion is extremenly slow for debug builds. If you want to generate more than 10k particles doing a release build is mandatory.

Because the crate has link time optimizations enabled the final build step may be noticeably slow.

> cargo build --release

Running

Builds will be in the ./target folder.

> target/release/rdla render --display

Cloud Rendering

“Hold on to you lug nuts it’s time for an overhaul!”

– Stanley Ipkiss, ‘The Mask’.

To render in the cloud register at 3delight.com, open the 3Delight Display app and log in with your credentials.

Once you have done this all that is needed to launch a render in the cloud is to add the --cloud flag.

> target/release/rdla render --cloud --display

Command Line Interface

The app creates & renders point clouds based on diffusion limited aggregation.

USAGE:
    rdla [OPTIONS] [SUBCOMMAND]

FLAGS:
    -h, --help       Prints help information
    -V, --version    Prints version information

OPTIONS:
    -c, --config <FILE>    Sets a custom config file (default: ./rdla.toml)
    -p, --particles <N>    Number of particles to generate (default: 1000)

SUBCOMMANDS:
    dump      Dump the result into an .nsi stream or into a Stanford .ply file
    help      Prints this message or the help of the given subcommand(s)
    render    Render an image of result with 3Delight

dump Subcommand

Dump the result into an .nsi stream or into a Stanford .ply file.

USAGE:
    rdla dump <FILE>

FLAGS:
    -h, --help       Prints help information
    -V, --version    Prints version information

ARGS:
    <FILE>    File name of dump; extension determines format to use

render Subcommand

Render an image of the point cloud with 3Delight.

USAGE:
    rdla render [FLAGS] [FILE]

FLAGS:
        --cloud      Render using 3Delight Cloud
    -d, --display    Open a 3Delight Display
    -h, --help       Prints help information
    -V, --version    Prints version information

ARGS:
    <FILE>    Image file name to render to (default: out.exr)

Config File

The app looks for a config file named rdla.toml in the current directory.

This can be overridden with the --config flag.

[aggregation]
    show_progress = true
    random_seed = 42
    particles = 10000
    # Spacing can be changed over the iteration.
    # The 1st value is used for the first particle
    # placed and the last for the last particle.
    # In between, spacing is linearly interpolated.
    spacing = [1.0, 1.0]
    attraction_distance = 3.0
    repulsion_distance = 1.0
    stubbornness = 0
    stickiness = 1.0

    [aggregation.start_shape]
        # Try using "ring".
        shape = "point"
        # "ring" diameter.
        diameter = 0
        # No. of particles on
        # start shape.
        particles = 1

[particle]
    # Scale can be changed over the iteration.
    # The 1st value is used for the first particle
    # placed and the last for the last particle.
    # In between,scale is linearly interpolated.
    scale = [2.0, 2.0]
    # A wavefront OBJ (converted to triangles for now)
    # to instace instead of a sphere particle.
    instance_geo = "" # "assets/cube.obj"
    subdivision = true

[material]
    color = [0.5, 0.6, 0.8]
    roughness = 0.3
    metallic = 1.0
    specular_level = 0.8

[environment]
    texture = "assets/artist_workshop_1k.tdl"
    intensity = 1.2

[nsi_render]
    # Images are square
    resolution = 2048
    shading_samples = 100
    oversampling = 20
    bucket_order = "circle"

    [nsi_render.output]
        file_name = "out.exr"
        display = false