[gst-devel] preroll (was: Re: [gst-cvs] thomasvs gstreamer: gstreamer/docs/random/ds/)

Thomas Vander Stichele thomas at apestaart.org
Wed Jul 14 04:41:22 CEST 2004


On Wed, 2004-07-14 at 12:50, Benjamin Otte wrote:
> On Wed, 14 Jul 2004, Thomas Vander Stichele wrote:
> 
> > Hm, it's true that it doesn't state this, and I'm not sure it should
> > either.  There's no reason why a bin cannot override the vfunction so
> > that instead of putting it in itself, it puts it into a second bin that
> > is private to the given bin.  Why would you do this ? To make a bin that
> > can preroll would be one example; the private bin can be set to playing
> > to make sure everything is autoplugged, while the public bin that is put
> > in the app pipeline is still in paused.  This is one of the things that
> > currently is not done that Jorn complains about.
> >
> Speaking of preroll:
> Do we have a benchmark or other demonstartion that illustrates the need of
> prerolling?
> 
> The only justification for preroll I keep hearing is that Jorn had issues
> (iirc that was slowly pausing pipelines, which is not even an issue with
> preroll, but whatever). And the only hard benchmarking data I have tells
> me that mp3 pipelines need 0.02 seconds before they start playback and
> vorbis audio 0.07 seconds. (tested with gstreamer/tests/spidey_bench
> file.ext) And I don't think those times are an issue. So it sounds a lot
> like premature optimization from my point of view.

Wasn't it you that said your harddrive takes a few secs to spin up when
a track starts playing ? So doesn't that mean that in any case, going
from paused to playing takes that amount of secs ? Am I missing
something ?

I still see similar issues in rhythmbox as well, and switching between
paused and playing still isn't as zippy as it ever was in xmms.

I'm not really sure what other methods are currently possible to solve
this.

Anyways, whether or not preroll makes sense is beside the point - the
architecture should somehow support a concept like it, since obviously
other frameworks have decided it is a useful concept and not a premature
optimization.

Thomas


Dave/Dina : future TV today ! - http://www.davedina.org/
<-*- thomas (dot) apestaart (dot) org -*->
I can't go away with you on a rock climbing weekend
What if something's on TV and it's never shown again
Just as well I'm not invited I'm afraid of heights
I lied about being the outdoor type
<-*- thomas (at) apestaart (dot) org -*->
URGent, best radio on the net - 24/7 ! - http://urgent.fm/






More information about the gstreamer-devel mailing list