[mythtv-users] Heh - will be nice to have such FE - isn't?

Jeff Siddall news at siddall.name
Tue Jun 11 14:11:48 UTC 2013


On 06/11/2013 10:00 AM, Joseph Fry wrote:
>
>             This device is actually very good.
>             The API for h264/mpeg decode is the v4l2 API in linux. It
>             looks like Samsung
>             supplied patches and drove some of the changes for v4l2.
>             Theres a git repo
>             http://git.infradead.org/__users/kmpark/public-apps/tree/__HEAD:/v4l2-mfc-example
>             <http://git.infradead.org/users/kmpark/public-apps/tree/HEAD:/v4l2-mfc-example>
>             that provides a sample program showing how to use the v4l2
>             api for decoding!
>
>             If this was build in to MythTV then I think it would make a
>             very nice
>             frontend.
>             Of course you'd have to consider the deinterlacing, but a
>             quad core
>             1.7GHz should be
>             capable, right.
>
>
>         Quad core ARM?  Umm... not likely.  ARM != i7
>
>
>     For deinterlacing? the decode will be HW accelerated.
>
>
> Depends upon the deinterlacing used. Assuming that you can use the CPU
> to deinterlace a GPU decoded video (not sure it it will even let you do
> that), you would have a range of deinterlacers to choose from.
>
> Simple deinterlacers like One Field, or Bob, use minimal CPU... they do
> little processing of the frames.  Advanced deinterlacers actually detect
> motion in the frames and manipulate the fields to make each frame appear
> as close to complete as possible.  The motion detection, and
> manipulation of the feilds can be VERY cpu intensive when your
> deinterlacing HD video in realtime.

Correct.  I was expressing concern with the ability of an ARM to do 
advanced deinterlacing.


More information about the mythtv-users mailing list