Re: WebGPU Compute Example

Yes you’re right, it was more a proof-of-concept than anything else to demonstrate geometry generation via a compute pipeline feeding into a render pipeline and no longer matches the IDL file. (Also the Metal shaders are a bit of a giveaway :P)

When a proper WebGPU implementation comes along I’ll re-write it :)

Sent from my iPhone

On 1 Mar 2019, at 6:49 am, Kai Ninomiya <kainino@google.com<mailto:kainino@google.com>> wrote:

Thanks for the clarification, Dean!

Jerran: FYI, for more info on the high-level history of things around WebGPU, check out the Wikipedia article<https://en.wikipedia.org/wiki/WebGPU>.

On Thu, Feb 28, 2019 at 11:31 AM Dean Jackson <dino@apple.com<mailto:dino@apple.com>> wrote:


On 28 Feb 2019, at 10:12, Jerran Schmidt <jerran.schmidt@autodesk.com<mailto:jerran.schmidt@autodesk.com>> wrote:

I’ve been following the development of WebGPU mainly from a compute perspective and coming from an OpenGL (and now Vulkan) background I’ve found the API itself very easy and intuitive to use.
I couldn’t find any samples on using WebGPU for compute at the time that I put my first simple example on github but I’ve since experimented with going a little further and implemented marching cubes:
Pages:
https://jerrans.github.io/webgpu-marching-cubes/


Code:
https://github.com/jerrans/webgpu-marching-cubes


I didn’t have any issues other than whether setting the threads per threadgroup is something that would be exposed or not.

In any case the standard looks to be progressing well and I’m hoping to be able to contribute in the future.

To be clear, what you've used there is Apple's initial WebGPU proposal, which we've since renamed WebMetal, and will soon be removed from Safari completely - once we have a real WebGPU implementation.

This will require a fairly significant rewrite of your example, although the fundamentals are approximately the same.

Dean