On Fri 30 August 2013 09:28:36 Pawel Osciak wrote:
> On Fri, Aug 30, 2013 at 3:42 PM, Hans Verkuil <hverkuil@xxxxxxxxx> wrote:
>
> > On 08/30/2013 04:17 AM, Pawel Osciak wrote:
> > > Add bits for previous, golden and altref frame types.
> > >
> > > Signed-off-by: Pawel Osciak <posciak@xxxxxxxxxxxx>
> >
> > Kamil, is this something that applies as well to your MFC driver?
> >
> > > ---
> > > include/uapi/linux/videodev2.h | 4 ++++
> > > 1 file changed, 4 insertions(+)
> > >
> > > diff --git a/include/uapi/linux/videodev2.h
> > b/include/uapi/linux/videodev2.h
> > > index 437f1b0..c011ee0 100644
> > > --- a/include/uapi/linux/videodev2.h
> > > +++ b/include/uapi/linux/videodev2.h
> > > @@ -687,6 +687,10 @@ struct v4l2_buffer {
> > > #define V4L2_BUF_FLAG_TIMESTAMP_UNKNOWN 0x0000
> > > #define V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC 0x2000
> > > #define V4L2_BUF_FLAG_TIMESTAMP_COPY 0x4000
> > > +/* VP8 special frames */
> > > +#define V4L2_BUF_FLAG_PREV_FRAME 0x10000 /* VP8 prev frame
> > */
> > > +#define V4L2_BUF_FLAG_GOLDEN_FRAME 0x20000 /* VP8 golden
> > frame */
> > > +#define V4L2_BUF_FLAG_ALTREF_FRAME 0x40000 /* VP8 altref
> > frame */
> >
> > Would it be an idea to use the same bit values as for
> > KEYFRAME/PFRAME/BFRAME?
> > After all, these can never be used at the same time. I'm a bit worried
> > that the
> > bits in this field are eventually all used up by different encoder flags.
> >
>
> VP8 also has a concept of I and P frames and they are orthogonal to the
> concept of Prev, Golden and Alt. There is no relationship at all, i.e. we
> can't infer I P from Prev Golden Alt and vice versa.
Ah, OK. Me no VP8 expert :-)
> For the I, P
> dimension, we need one bit, but same could be said about H264, we need 2
> bits, not 3 there (if it's not I or P, it's B), and we have 3 bits
> nevertheless.
> For Prev Golden Alt dimension, we do need 3 bits.
> Now, technically, in VP8, the first bit of the frame header indicates if
> the frame is I or P, so we could technically use that in userspace and
> overload I P B to mean Prev Golden Alt for VP8. But while I understand the
> problem with running out of bits very well, as I and P exist in VP8 as
> well, it would be pretty confusing, so it's a trade-off that should be
> carefully weighted.
Are prev/golden/altref frames mutually exclusive? If so, then perhaps we
should use a two-bit mask instead of three bits. And those two-bits can
later be expanded to more to support codecs that have more than four
different frame types.
Regards,
Hans
>
> Regards,
> Pawel
>
>
> > Regards,
> >
> > Hans
> >
> > >
> > > /**
> > > * struct v4l2_exportbuffer - export of video buffer as DMABUF file
> > descriptor
> > >
> >
> >
>
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
