[FFmpeg-devel] [PATCH] Add hwaccel_data infrastructure

[Michael Niedermayer]

On Tue, Mar 03, 2009 at 11:12:32PM +0100, Gwenole Beauchesne wrote:
> Hi,
> 
> Le 3 mars 09 ? 22:37, Michael Niedermayer a ?crit :
> 
> > On Tue, Mar 03, 2009 at 01:29:17PM +0100, Gwenole Beauchesne wrote:
> >> Hi,
> >>
> >> This patch implements the suggested hwaccel_data infrastructure.
> >>
> >> - hwaccel_data is allocated by the player and matches what we  
> >> currently
> >> name vdpau_render_state or vaapi_render_state. The aim is to  
> >> replace the
> >> use of AVCodecContext::data[0] (reserved for YV12 pixels -- future  
> >> use).
> >
> > is there an advantage over using data[3] ?
> 
> I don't see any actually. In either case, data[3] (or hwaccel_data)  
> needs to be filled in with e.g. mpi->priv in the mplayer case for  
> example.
> 
> >> - hwaccel_data_private is optional (check  
> >> AVHWAccel::priv_data_size) and
> >> allocated by FFmpeg. This private data has the lifetime of the  
> >> decoded
> >> AVFrame. It's deallocated through the ::close() function. I don't  
> >> really
> >> like the name. For VA API, I implemented it as  
> >> ff_vaapi_destroy_picture().
> >
> > this looks asymetric
> > alloc by av_mallocz()
> > close by callback
> > why?
> 
> Because I suggested open/close and you said only when needed. :) So I  
> assumed it was the wrong route since, currently, VA API only needs  
> av_mallocz() at frame buffer creation and do extra work at frame  
> buffer destruction.

can you explain what this extra work is?

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

If you really think that XML is the answer, then you definitly missunderstood
the question -- Attila Kinali
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20090304/cd4bef20/attachment.pgp>