On Sat, 3 Feb 2007, Manu Abraham wrote:
On 2/3/07, Hans Verkuil <hverkuil@xxxxxxxxx> wrote:
On Saturday 03 February 2007 19:23, Manu Abraham wrote:
> On 2/3/07, Hans Verkuil <hverkuil@xxxxxxxxx> wrote:
> > On Saturday 03 February 2007 18:53, Manu Abraham wrote:
> > > On 2/3/07, Hans Verkuil <hverkuil@xxxxxxxxx> wrote:
> > > > RFC MPEG encoding and decoding V4L2 API additions
> > > > Version 0.1
> > >
> > > Wouldn't it be a bit overkill to add one more decoder API into
> > > the kernel ?
> >
> > "decoding V4L2 API additions". Not a new API, just additions to be
> > able to fully utilize MPEG hardware decoders. If you have good
> > alternatives that can be used instead of the proposed additions
> > I'll be glad to hear them.
>
> Currently, we have MPEG decoders using video.h/osd.h
> Almost all DVB capable applications use them as well.
Currently, it seems that the only driver at kernel using DVB mpeg ioctls
is av7110. This hardware seemed to be discontinued by Texas, and is
lacking a proper maintanership (there are some pending bugs on bugzilla
waiting for someone to deal with).
> IMHO, it would be probably a pain for applications supporting
> multiple API 's doing the same thing.
Hmm, I'll have to look at that more closely tomorrow. I wasn't aware
that the dvb API had support for decoding.
Are there any existing v4l2 drivers using parts of the dvb API? I'm not
certain how well they mix.
Currently, the AV7110 uses some parts of V4L2 for the analog stuff.
These were based on the old Siemens cards, later on they were
manufactured by a plethora of vendors (Technotrend, Hauppauge, Galaxis
and many others ) making it very popular. These cards are popularly
knowns as the AV7110 based Full Fledged (FF) cards.
If the mpeg implementation on DVB API is generic enough to work with
non-av7110 hardware, we should use those ioctls for ivtv. Otherwise, we
should work in a way that won't result on userspace breakage when adding
newer stuff to the ioctl. In the latter case, as on DVB, there are no
reserved space on current structs for API additions, such change will
likely result on defining a newer ioctl number with the old struct fields
(or a more generic approach) plus the new required ones.
I think some newer devices also will be based on this API, the
Decypher [1] H.264 decoders from Micronas. Still work is going on with
that device, AFAIK.
We should analyse the requirements of those devices and if the API
provides enough support. Is there anybody already working on such
hardware?
Cheers,
Mauro Carvalho Chehab
http://linuxtv.org
mchehab@xxxxxxxxxxxxx
_______________________________________________
linux-dvb mailing list
linux-dvb@xxxxxxxxxxx
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
