[gst-devel] empty buffers & GST_BUFFER_FLAG_GAP

[Jan Schmidt]

On Wed, 2006-01-18 at 21:00 +0100, Stefan Kost wrote:
> Hi Jan & Wim,
> Jan Schmidt wrote:
> > On Wed, 2006-01-18 at 17:36 +0100, Stefan Kost wrote:
> > 
> > 
> >>I think there are use-cases for both applications. I'd like to read your design 
> >>document and then continue the discussion.
> > 
> > 
> > I just added a first draft of the idea in
> > docs/design/part-sparsestreams.txt (in core)
> > 
> I've just read
> http://cvs.freedesktop.org/gstreamer/gstreamer/docs/design/part-sparsestreams.txt?view=markup
> 
> Thanks for adding my use case to it. Still I don't see how it will work. Imagine 
> the following pipeline:
> 
> simsyn ! echo ! audiosink
> 
> simsyn is somthing like audiotestsrc, but instead of a freqeuncy you give it a 
> musical note (c-3). The note will be set via GstController from time to time. 
> Setting a note also triggers simsyn volume envelope. In the _create() function 
> simsyn checks the envelope level before rendering a buffer. Whenever the level 
> go close to zero it just create silent buffers (avoid calculating sines etc.).
> 
> Next in the pipeline is the (imaginary) echo element. An echo is usaly 
> implemented by using a ringbuffer, where data output-data gets feedback into the 
> ringbuffer. This produces decaying echos.
> 
> When simsyn determines it produces silence, I still want echo to be called, as 
> echo still has data to produce. When echo finds out that the output level from 
> the ring-buffer goes below zero it can mark output as silent too and remember 
> the state. If next silent buffer are comming, it can just output silence too.
> 
> I don't know if this can be done with the technique you've described in the 
> part-sparsestreams.txt. Unfortunately I don't know about the 'SCR in MPEG data' 
> so I can't map the 'for audio, 3) is the same case as in 1)'.

In MPEG, SCR just represents the 'current time' of the stream. In your case, once the source knows that
it will be producing silence until time 'n', it can update the start
time of the segment to time 'n'. The echo element can handle that by
outputting all data until that time. 

Hopefully that makes it clearer? I updated the sparsestreams.txt
slightly to try and clarify that point.

J.

-- 
Jan Schmidt thaytan at noraisin.net

<MikeS> thomas is pretty anal about changes now, though. He's allowed to
rearrange the entire cvs repository in bizarre ways, but other people
aren't even allowed to fix crasher bugs unless they can write testcases
for them.