listserver  

Re: [OzSilverlight] RE: Hardware accelerated video?

David Connors
Thu, 30 Oct 2008 01:44:33 -0700

Scott,

You probably think I'm yanking your chain and I know this is an SL list but
seriously "I wouldn't be discouraged by Flash 10's new toys as in the
end..Flash is still Flash :D" isn't really a good way to advocate a
platform.  Props where props are due - Flash 10 does look pretty good.

Other replies inline...

On Thu, Oct 30, 2008 at 3:23 PM, Scott Barnes <[EMAIL PROTECTED]>wrote:

>  Never stated hardware acceleration was a bad thing? You asked or assumed
> I implied that but I thought I rejected that and answered your question that
> it has to do with implementation.
>
So to clarify what are the parts of the Adobe implementation that are bad
that make you think the jury is out?

> As for hardware acceleration in the runtime, in what way is that going to
> be beneficial overall (i.e. what are you expecting to see or view in this
> case?).
>
I dunno - I'd be expecting to see all of the normal sorts of eye candy
enabled by using dedicated hardware:
1. High quality filtering and resampling  - not computationally feasible in
software
2. Alpha transparency of video   - not computationally feasible in software
3. Transformation of video while playing   - not computationally feasible in
software
4. 3D with trilinear filtering so you can read text that has come out of the
other end of the rendering pipeline. NFI if Flash 10 does that (probably
not) but either way you're never going to do it in software.

etc

> Cost has to do with Streaming not so much the client viewing, all those
> bytes add up and someone has to pay the bill, especially with CDN folks J
>
What bearing does the hardware acceleration model on the client have on the
kbps of a video stream it is downloading?

> Pre-computation? Could you elaborate on what you mean there?
>
I probably used a really poor example here that will confuse the
conversation - but say I open a massive image in PS11/CS4. You can
(relatively) smoothly zoom into it to your hearts content in real time
because the scaling is retargetted to hardware. In that case their use of
hardware acceleration (while not at all relevant to the discussion at hand)
appears to be very good. I'd presume that if they can enable scenarios like
that in an app as old and as complex as PS that they'd do a relatively good
job of using it in a client runtime. Note: I am not saying that downloading
a massive image and processing it on the client is a substitute for DeepZoom
... I just used that as an app that does something similar (hence I said bad
example). The point of pre-computation is that when SL is displaying a Deep
Zoom image it is only really working with screen resolution or something
pretty close.

I would be curious to know how it would go, performance wise, with a 30 meg
source image for example.

> I won't enter the debate around Sony vs. Samsung,
>
LOL. Okay ... but you're the one who brought up the analogy.

> it's more towards the analogy than the depth of which two brands approach
> consumers in which ways. The fact here in the US, is that you walk into any
> best buy store, look at the TV's on full view and the decision comparisons
> around which to buy typically live in and around price, chrome presentation
> and warranties. It's extremely hard to spot the difference between each HD
> TV on display, there are certain tricks applied – like playing Finding Nemo
> with various tweeks to color profiles to give a "brighter vs. darker" visual
> queue that somehow one trumps the other.
>
> So meh :0
>
What was the point of your analogy? Are you saying that Flash is all tricks
but SL is the real deal? Or that they're all the same and you can't tell? I
don't understand?

-- 
David Connors ([EMAIL PROTECTED])
Software Engineer
Codify Pty Ltd - www.codify.com
Phone: +61 (7) 3210 6268 | Facsimile: +61 (7) 3210 6269 | Mobile: +61 417
189 363
Address Info: http://www.codify.com/contact