#shader #graphics #game #derive-debug #debugging

macro crabslab-derive

Slab allocator focused on GPU compute (rust-gpu)

10 unstable releases (3 breaking)

0.4.2 May 20, 2024
0.4.1 May 14, 2024
0.4.0 Feb 28, 2024
0.3.0 Feb 7, 2024
0.1.3 Jan 5, 2024

#117 in #derive-debug

22 downloads per month
Used in 3 crates (via crabslab)

MIT/Apache

21KB
430 lines

slabcraft for crabs

What

crabslab is a slab implementation focused on marshalling data between CPUs and GPUs.

See the example below.

But Why?

It's hard to get data onto GPUs in the form you expect.

To marshall your data correctly you must know about the alignment and sizes of the underlying representation of your data. This will often surprise you!

Working with a slab on the other hand, only requires that your types can be written into an array and read from an array.

Opinion

Working with shaders is much easier using a slab.

Shader code can be written in Rust with rust-gpu, which will enable you to use this crate on both CPU and GPU code.

rust-gpu

This crate was made to work with rust-gpu. Specifically, with this crate it is possible to pack your types into a buffer on the CPU and then read your types from the slab on the GPU (in Rust).

Other no-std platforms

Even though this crate was written with rust-gpu in mind, it should work in other no-std contexts.

And How

The idea is simple - crabslab helps you manage a heap of contiguous u32s (roughly in the form of Vec<u32>). Types implement the trait SlabItem which writes the type into an index of the slab as contiguous u32s and also reads them out symmetrically.

crabslab includes:

  • a few traits:
    • Slab
    • GrowableSlab
    • SlabItem
  • a derive macro for SlabItem for your types
  • a few new structs for working with slabs
    • Id
    • Array
    • Offset
  • a helper struct CpuSlab which wraps anything implementing GrowableSlab
  • feature for deriving SlabItem for glam types

Example

use crabslab::{CpuSlab, Slab, GrowableSlab, SlabItem, Id};
use glam::{Vec3, Vec4};

#[derive(Debug, Default, SlabItem, PartialEq)]
struct Light {
    direction: Vec3,
    color: Vec4,
    inner_cutoff: f32,
    outer_cutoff: f32,
    is_on: bool
}

impl Light {
    fn standard() -> Self {
        Light {
            direction: Vec3::NEG_Z, // pointing down
            color: Vec4::ONE, // white
            inner_cutoff: 0.5,
            outer_cutoff: 2.6,
            is_on: true
        }
    }
}

fn cpu_code() -> (Id<Light>, Vec<u32>) {
    let light = Light::standard();
    // Create a new slab on the CPU-side.
    // Using CpuSlab make `append` unambiguous, as `Vec` has its own `append` function.
    let mut slab = CpuSlab::new(vec![]);
    let id = slab.append(&light);
    (id, slab.into_inner())
}

fn shader_code(light_id: Id<Light>, slab: &[u32]) {
    let light = slab.read(light_id);
    assert_eq!(Light::standard(), light);
}

let (light_id, slab) = cpu_code();
// marshalling your data depends on which GPU library you are using...
shader_code(light_id, &slab);

Dependencies

~260–700KB
~17K SLoC