"but potentially a lot more complex."

I don't think so.
I don't see where is the complexity actually.

You can process your vertex in your computeShader the same way you would do it in a vertexShader.

From the computeShader, you receive a buffer with the result. Use it as input of a renderPipeline and apply zero computation in your vertexShader

//in the vertexShader
output.position = bufferFromComputeShader;

Even if you use the computed-buffer directly in a renderPipeline, you can use a staging buffer as an intermediate step in order to access to the data on the CPU side for the debugger-side.

It's honnestly a bit complex to do with vanilla webgpu but it's easy to do with my library XGPU :)

Look at the sample "Video3D" .
https://xgpu-samples.netlify.app/samples/Video3D

I processed the vertex in the computeShader (in this particular case, my renderPipeline draw points not triangles so it's a bit différent but the core idea is the same)

In order to access to the data on the CPU side, you just need to write

 buffer.onOutputData = (data) => {
      
 }

it couldn't be easier :)
(I tryed a minute ago , it works - look at the line 115 in the sample - )

Look at the sample "ComputeShaderCanvas2D" to have a better understanding on how to receive and use data that come from a ComputeShader. I tryed to get it as readable and easy-to-use as possible.

EDIT : I never tryed to do a computation (in a computeShader) with GPUBuffer.stepMode = "vertex" (it's required to do what you want) but I think it should work. I'll try :)

EDIT2 : More I think about it, more I think it's not that easy.... I'm giving a try !

[Feature Request] Vertex Shader storage read/write #4355

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions