Worley 2D

[vladinator1000]

Heya @johanhelsing , I love this crate! I'm working on a Worley implementation, it's almost there but I have some strange issues. This is the rendered output of this example:

$env:WGPU_BACKEND="dx12"; cargo run --example worley

I'm developing on windows and added the wgpu backend env var to reduce console warnings.

I used this repo for inspiration: https://github.com/bevy-interstellar/wgsl_noise/blob/main/worley_vec2f.wgsl

Do you see where I'm making a mistake here? Looks like it doesn't handle negative coordinates well, because when x and y are positive (top right) there are no visual issues.

[vladinator1000]

Turns out glam's fract implementation is different: https://docs.rs/glam/latest/glam/f32/struct.Vec3.html#method.fract

I used fract_gl and it worked! Should be good for review now @johanhelsing.

On the CPU:

The only thing that bothers me is that we use value.y in the CPU demo it looks quite funky. Is it supposed to do that?

The shader example looks pretty cool though:

[vladinator1000]

Review notes

[vladinator1000]

Added mouse input to the example

2025-01-14.17-41-42.mp4

[johanhelsing]

Looks great, sorry for the slow response.

Did you check if it it produces similarish values on the CPU and GPU?

[vladinator1000]

Thanks @johanhelsing, addressed your comments, but the CPU and GPU values look different and I'm not sure how to fix that.

CPU version $env:WGPU_BACKEND="dx12"; cargo run --example worley

GPU $env:WGPU_BACKEND="dx12"; cargo run --example worley_shader

[johanhelsing]

It's a bit unfortunate that we break the promise that gpu and cpu-noise are the same... It would be awesome if we could figure out what's causing the difference.

[johanhelsing]

I pretty closely at the two versions side-by-side, but couldn't find anything obvious causing the discrepancies :/ Guess maybe we'll have to go line-by-line and figure out what causes it

[johanhelsing]

Turns out they're quite similar after all. I modified the examples quite heavily to make sure the inputs were the same, and they look similar.

[johanhelsing]

Cleaned up the changes and made PR to your branch: vladinator1000#1

Merge that and revert the changes that break the doc links, and this should be good to go :)

[vladinator1000]

Wow, thanks @johanhelsing, turns out it was almost there ❤️ I fixed the docs.

[johanhelsing]

Thanks! This is a really cool addition. Thanks for sticking around even though my responses have been far in-between!