[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Public WebGL] WebGPU



On Wed, Feb 8, 2017, 5:52 AM Kirill Dmitrenko <dmikis@yandex-team.ru> wrote:
It is without a doubt an interesting undertaking. However I wonder how much performance could such tech really give comparing with, let's say, Vulkan backed WebGL 2 (or 3, or some kind of 2.X)? Will there be an advantage over WebGL extended with GPGPU capabilities?

WebGL retains all of the design-related performance problems of OpenGL even if it is implemented on Vulkan (which will become feasible, with ANGLE adding a Vulkan backend). It would be faster (reducing driver validation overhead) in some cases, but not nearly as fast as a new API (which also reduces our validation overhead)

Another difficulty is that in order to bring compute capabilities to WebGL on Apple platforms, someone needs to write an entire OpenGL ES 3.1 implementation on top of Metal. This could be possible but it's a huge undertaking. A new web API designed to fit all of the new native APIs closely would be much easier.

There's one thing that could give WebGPU a huge advantage over WebGL: an ability to share a context between several workers to construct command queue concurrently (as it is in Metal and, AFAIK, in D3D12 and Vulkan). That would be awesome.

This is definitely on the radar.

-Kai

08.02.2017, 01:34, "Andrew" <grizzly33@gmail.com>:
> Just found this, and since there seems to be no thread about it yet, I wonder what everyone here thinks about it:
>
> https://webkit.org/blog/7380/next-generation-3d-graphics-on-the-web/


-- 
Kirill Dmitrenko
Yandex Maps Team

-----------------------------------------------------------
You are currently subscribed to public_webgl@khronos.org.
To unsubscribe, send an email to majordomo@khronos.org with
the following command in the body of your email:
unsubscribe public_webgl
-----------------------------------------------------------

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature