[whatwg] video tag javascript library for contemporary browsers

On 10/15/08, Chris Double <chris.double at double.co.nz> wrote:
> On Thu, Oct 16, 2008 at 7:35 AM, Michael A. Puls II
> <shadow2531 at gmail.com> wrote:
>> I *think* it has to do with the lack of hardware acceleration (even in
>> webkit's implementation). It seems like it's all CPU driving the video
>> element. No beefy CPU, no usable video element.
>
> Yes, this is certainly an issue. A player using hardware acceleration
> will outperform a player that doesn't. You won't be able to do things
> like overlay HTML over the plugin area, perform effects and
> transformations, copy the image of the video frame to canvas, etc with
> the plugin as a result.

Maybe <video> needs something (currently. maybe not in a few years)
like a wmode param where the author can suggest (and the user can
ultimately override if necessary) whether it's accelerated or not.

Then, a site might provide <video wmode="something"> all  by itself on
a separate page as an alternative to the normal
all-cpu-driven-can-be-overlayed  version.

Would be something neat to play with to see if it helps. Something
like video.mozWmode.

>> But, I don't know details. Just know that the videolan
>> plugin can play theora videos with very little cpu usage, while the
>> *experimental* <video> implementations use 100% cpu, display video at
>> like 2fps and play audio like crap, unless you have a fast computer
>> where you can't notice.
>
> Can you provide details of the specs of the computer, operating
> system, and the page that you see these issues so I can test and fix
> any issues?

Sure, no problem.

Example hardware that *needs* the acceleration that VideoLan (and
whatever flash does) provides:

PIII 700MHz + 256MB ram + ATI Rage Mobility + 5400RPM hard drive + ESS Maestro

PII 350MHz + 192MB ram + AGP TNT2 + 5400RPM ATA33 hard drive + Diamond
sonic impact s90

PIII 733MHz + 320MB ram + PCI GeForce2 MX400 + ata100 7200RPM har
drive + Sound blaster audigy

1.3GHz P4 + 256MB PC800 Rambus + ATA100 7200RPM + AGP TNT2 (better,
but not perfect) + Sound Blaster live

(WinXP, 8Mb net connection, lastest drivers available)

To sum it up, if you have a P4 or below, <video> requires something
like 1.6GHz (or way higher if ads are on a page I assume) or so, while
a PII 350Mhz does well with the videolan plugin.

<http://people.opera.com/howcome/2007/video/controls.html> (bad)
<http://people.opera.com/howcome/2007/video/wikipedia/macaw.html>
(absolutely horrible)
<http://tinyvid.tv/show/25tuo4iijnin8> (horrible)
<http://tinyvid.tv/show/3m2v1wnodj23i> (even worse. Stops playing if
Firefox is focused)

In the case of Firefox, <video> performance is less bad if Firefox
isn't focused (or if Firefox is done fetching the file). However, even
then, it's still using 100% cpu and typing in any other application
causes the audio to studder big time.

Even when a <video> is paused, Firefox uses 100% cpu. The same type of
results are present in webkit and Opera's experimental video builds,
so it's not only Firefox.

-- 
Michael

Received on Wednesday, 15 October 2008 21:03:28 UTC