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

Re: [Public WebGL] Is section 6.3 still needed ? (was: New Rendering Pipeline ?)



Cedric Vivier wrote:
> With the news of Chrome jumping into the GPU-accelerated HTML
> compositing train as well, I stumbled upon an interesting document
> about how Chrome handles accelerated compositing, WebGL content and
> NaCl 3D rendering within the same "GPU process" :
> https://sites.google.com/a/chromium.org/dev/developers/design-documents/gpu-accelerated-compositing-in-chrome
Woohoo!

Compositing is taking something like 40% to 85% of the frame time in my
app (at high screen resolutions - and depending on OS, browser and
hardware).  Doing it using the GPU should get that down to the roughly
5% that it should be and result in a doubling to quadrupling of frame
rates!  That is a MAJOR win.  If I can spend 4x longer rendering content
then that's is literally the difference between something that looks
like Mario64 and something that looks like Red Dead Redemption!  I
strongly encourage all of the browser writers to do it as a matter of
priority.  However, it's only going to be such a huge win if the image
that WebGL just rendered STAYS IN THE GPU.  If you have to pull it back
into main memory and push it back into the GPU later for reasons of
threading or who-knows-what-else - then you just blew away most of that
benefit.  If we really want to see cutting-edge interactive 3D on the
web - this is a "must have" thing.

Concerns about the reliability of graphics drivers should be pretty
minimal - I work with 3D all the time - I've been doing it for 20 years,
it's my job.  Over the past 4 or 5 years, I've seen very few driver
problems - and they've mostly been of the kind that only affects very
obscure corners of the architecture.  When you have WindowsVista/Win7
doing their window compositing using the GPU, you know it's going to be
pretty solid.

  -- Steve

-----------------------------------------------------------
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: