Here is google's example to test new syntax https://gist.github.com/danteissaias/b855d765996ef46b8a5b05d0fa62db76

Currently there is an issue with webidl.converters.GPUBindGroupLayoutEntry which prevents createBindGroup from working.

whats up with the typings?

I copied the latest ones from the repo was unsure of whether to keep in the documentation.

which repo?

https://github.com/gpuweb/gpuweb

There was a commit ID in a comment references this repo.

ah, we always handroll our webgpu typings

Ok, I’ll do that tomorrow.

Could you investigate the error with createBindGroup?

yes, i will do that

There was a commit ID in a comment references this repo.

ah we havent updated in a while (or rather, since the first few PRs). that commit used to refer to the commit of the webidl that the typings match to, but i think its better we remove it as we have an idl file stored alongside the code

@danteissaias fixed the createBindGroup problem, now the result I am getting is an typedarray with 0s

Somethings going wrong…. Will have to investigate.

@danteissaias
manually validated the wgsl with the same naga version that wgpu is using:

error: unknown attribute: 'stage'
   ┌─ test.wgsl:10:2
   │
10 │ @stage(compute) @workgroup_size(8, 8)
   │  ^^^^^ unknown attribute

Could not parse WGSL

So either google’s example is outdated to the spec or wgpu-rs is…

https://www.w3.org/TR/WGSL/#entry-point-decl looks like the example is outdated. Try with this syntax.

yep, was just going through the spec as well, thank god its not a problem with wgpu/naga 😅

hm no luck still, the shader validates now, but still getting 0s in typedarray

I updated the gist. Try with new shader changes.

interesting new output:

Vector component[0] type Scalar { kind: Float, width: 4 }
Float32Array(6) [ 0, 0, 0, 0, 0, 0 ]

https://gist.github.com/danteissaias/ba81b617ef1143a7ceb83e63f116c3c9

This is the hello-compute example updated with the newest shader syntax from the wgpu repo. It works. So must be an issue with google's shader.

👍