shader-types

Crates.ioshader-types
lib.rsshader-types
version0.2.2
sourcesrc
created_at2020-07-12 07:51:55.991128
updated_at2020-08-30 17:29:21.722263
descriptionVector and Matrix types that are properly aligned for use in std140 uniforms
homepage
repositoryhttps://github.com/BVE-Reborn/shader-types
max_upload_size
id264316
size41,092
Connor Fitzgerald (cwfitzgerald)

documentation

README

shader-types

GitHub Workflow Status Crates.io Documentation License

DO NOT USE THIS CRATE

There are fundamental flaws with this method of defining padding. This crate is kept live and not yanked for backwards compat only. For a crate that serves this exact purpose, look at the wonderful glsl-layout which doesn't suffer from the same problems.

Old Readme:

Vector and Matrix types that are properly aligned for use in std140 uniforms.

All the types in this library have the same alignment and size as the equivilant glsl type in the default mode (std140).

This fixes the padding within members of structs but padding between members needs to be minded. The types in padding are there to make this easier.

Vectors are constructable to/from an array of their underlying type. Matrices are constructable to/from both 1d and 2d arrays as well as an array of the underlying vector type. (eg. Mat2 can be constructed from [Vec2; 2])

Features

  • bytemuck all types implement Zeroable and Pod.
  • mint enable conversions to/from equivalent mint types.

Example

For the following glsl:

layout(set = 0, binding = 0) uniform Block {
    mat4 mvp;
    vec3 position;
    vec3 normal;
    vec2 uv;
    int constants[3];
};

This struct is rife with padding. However it's now easy to mind the padding:

use shader_types::{Vec2, Vec3, Mat4, ArrayMember};

// Definition
#[repr(C)]
#[derive(Copy, Clone)]
struct UniformBlock {
    mvp: Mat4, // 16 align + 64 size
    position: Vec3, // 16 align + 12 size
    normal: Vec3, // 16 align + 12 size
    uv: Vec2, // 8 align + 8 size
    constants: [ArrayMember<i32>; 3] // 3x 16 align + 4 size
}

fn generate_mvp() -> [f32; 16] {
    // ...
}

// Construction
let block = UniformBlock {
    // Anything that can be converted to a [f32; 16] or [[f32; 4]; 4] works
    mvp: Mat4::from(generate_mvp()),
    position: Vec3::new([0.0, 1.0, 2.0]), // `from` also works
    normal: Vec3::new([-2.0, 2.0, 3.0]),
    uv: Vec2::new([0.0, 1.0]),
    constants: [ArrayMember(0), ArrayMember(1), ArrayMember(2)]
};

// Supports bytemuck with the `bytemuck` feature
unsafe impl bytemuck::Zeroable for UniformBlock {}
unsafe impl bytemuck::Pod for UniformBlock {}

let block_u8: &[u8] = bytemuck::cast_slice(&[block]);

MSRV

Rust 1.34

License: MIT OR Apache-2.0 OR Zlib

Commit count: 24

cargo fmt