swoorup/wgsl-bindgen

Generate typesafe Rust bindings for wgsl shaders in wgpu

Hi there 👋

My name is Swoorup Joshi. I'm a passionate about distributed & concurrent systems, and making high traffic application scale.

While I look at the platform/runtime as a tool to achieve the goal, my personal weapons of choice include: Rust or F#.

println!("${firstName}[email protected]")

wgsl-bindgen's Issues

Wire up wgsl_minifier as an option.

There is an hastily written code here to convert naga module to strings, which contains wgsl_minifier.

It should be wired up as a WgslBindgenOption instead.

Tidy up the output

When looking at the output, it appears as if a bit all over and messy right now.

Perhaps we could move assertions, impls to their own module and re-exported by root (is it even required)? Somehow use rustfmt instead of prettyparse?

Moving the assertions, impls around, we will need to ensure the struct usages as fully qualified.

Feature to generate bindings for bevy

wgsl-bindgen is able to currently generate extra bind group layout entries that can be customized to user's liking which is possibly different to that of Wgpu api.

See https://github.com/Swoorup/wgsl-bindgen/blob/main/wgsl_bindgen/src/bindgen/options/bindings.rs

Would be nice to get this to a stage where everything can be customized and target framework can be specified by the user, for eg: bevy?

Add constant counting number of buffers

I got bitten by setting low Limits values by hand. It would be nice if code emitted by wgsl-bindgen included counters for the number of bind groups, uniforms, storage buffers and so on, to set the appropriate values in the Limits struct.

Provide `min_binding_size` when generating the Bind Group Layout Descriptor.

We should probably provide a min_binding_size for uniform buffer when generating the Bind Group Layout Descriptor.

wgpu::BindGroupLayoutEntry {
  binding: 0,
  visibility: wgpu::ShaderStages::VERTEX | wgpu::ShaderStages::FRAGMENT,
  ty: wgpu::BindingType::Buffer {
    ty: wgpu::BufferBindingType::Uniform,
    has_dynamic_offset: false,
    min_binding_size: NonZeroU64::new(std::mem::size_of::<gpu_data::BatchUniformBuffer>() as _),
  },
   count: None,
}

Expose SHADER_STRING in bindings

Thanks for the great crate! Maybe a niche issue, but I'm using this to handle all the naga_oil preprocessing, generate some structs that have the correct layout, but otherwise I need to fit this in with an existin WGPU codebase (https://github.com/tracel-ai/burn). For that I need access to the final generated SHADER_STRING, which atm ends up as private.

Just have locally modified this for now, happy to create a PR if that sounds good!