[FFmpeg-devel] [POC] SDL display lavfi sink

[Luca Barbato]

On 12/27/2010 12:52 AM, Michael Niedermayer wrote:
> libavdevice is part of the (de)muxer layer, you cannot implement libvo in there
> without considerable unneeded complexity.

To clarify why the demuxer layer doesn't fit:

-ao -> you'd like to keep the main loop here to easily use the sound
card clock.

-vo -> you want to avoid unnecessary memcopies around and possibly write
directly the slices being decoded on the offscreen memory and then blit
the whole.

you might also add that ffmpeg right now works not optimally already on
networked situation since there is already a situation similar to the
audio-out.

Rightly the same could be said from other captures since again having
the encoding/decoding loop driven by something different might be nicer
in some situations.

The problems somebody might want to tackle is to find a solution more
optimal that doesn't:

- add too many unnecessary queues and memcpy around
- make us lose packets or samples while encoding/decoding

lu

-- 

Luca Barbato
Gentoo/linux
http://dev.gentoo.org/~lu_zero