[FFmpeg-devel] [PATCH 0/2] VDA decoder for ffmpeg

Michael Niedermayer michaelni at gmx.at
Wed Aug 22 05:14:14 CEST 2012


On Wed, Aug 22, 2012 at 09:43:27AM +0800, Xidorn Quan wrote:
> On Wed, Aug 22, 2012 at 5:30 AM, Michael Niedermayer <michaelni at gmx.at>wrote:
> 
> >
> > > 1. Origin H.264 decoder never guarantees that HWAccel will be used.
> > >
> > >    For some cases, the decoder doesn't call get_format, and we will
> > >    have no chance to ask it to use HWAccel. If it happens,
> > >    the h264_vda decoder will automatically fall back to software
> > >    decoding without any warning or something like this.
> > >
> > >    I don't think it is a proper behavior for this decoder. It should
> > >    either succeed to initialize and then use VDA to decode, or fail
> > >    if it can't and let caller to fall back itself.
> > >
> > >    In fact, sometimes, VDA can decode a video but H.264 decoder
> > >    refuses to use it and do decoding itself.
> >
> > how can these problems be reproduced? / where can i find h264 samples
> > that show these problems ?
> >
> 
> Any H.264 video with bit depth other than 8bit will prevent
> the origin H.264 decoder from invoking get_format.

if you simply grep for get_format in h264.c you can see a switch()
for the various depths, adding get_format() calls to all depths
that VDA supports should be quite easy.


> 
> My testing shows that VDA can decode video with 10bit bit depth
> though it doesn't output completely correct pictures.

what exactly is wrong on these pictures ? is VDA ignoring the depth
and using 8bit buffers/references ?

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

It is what and why we do it that matters, not just one of them.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20120822/3d20dc62/attachment.asc>


More information about the ffmpeg-devel mailing list