[mythtv] Proposal for updatng ffmpeg in Mythtv 0.21

Matthew Grant matthewgrant5 at gmail.com
Mon Jun 29 23:24:35 UTC 2009


Hi Jean-Yves,

Sorry about the name.

Please look carefully through my ramblings, I tend to write what I am thinking.

On Mon, Jun 29, 2009 at 11:38 AM, Jean-Yves Avenard<jyavenard at gmail.com> wrote:
> Hi
>
> 2009/6/29 Matthew Grant <matthewgrant5 at gmail.com>:
>> Scenario is that the VGPDAU backport really isn't enough to do
>> everything properly.  Updating the MtyhTV 0.21 to the lastest FFMPEG
>> and adding LATM AC3 support would be a good idea.  I am running an
>> older AMD64x2 Shuttle with a 4400 CPU, 1MB cache per core, with an ATI
>> 9250 AGP card.  My trials with MythTV 0.21 + fixes + patches  from
>> Jacque Y Avenard's build of 0.21 indicate that the hardware is capable
>> of full 1080p decode, but that it is bottlenecking on the single
>> threaded H264 MP4 decode path in the FFMPEG libraries.
>
> Upgrading to ffmpeg trunk wouldn't fix your issue because
> multi-threaded decoding isn't in trunk yet but in a branch that has
> yet to be merged.
> just get a cheap nvidia card, will fix all your issue

If you look above, my hardware only has an AGP slot, and the Nvidia
hardware for AGP does not support the VGPDAU....  I don't see the
point in spending a lot of money for a new machine when my current
hardware is perfectly capable of performing adequate 1080p h264
decoding.  From my playing around I can tell it is almost there.

There is also the fact that the needs of lot of MythTV users here in
NZ and Norway are not being meet with the necessary h264 support for
their environments.  The DVB-T in these countries is bit odd, and is
some of the first in the world (MHEG EPG and LATM AAC audio).  0.22 is
not going to be out in the next 2-3 months, and it could easily be
12-18 months before it is available?

How soon will the stabilization for 0.22 take place?  I would be
willing to start working on testing this, and fixing UI bugs/crashes
if there was going to be something like a coding marathon to get most
of the clean up done.  --

I would like to branch 0.22 RC in svn to test and develop the
multi-threaded FFMPEG support as an alternate decode method to
standard ffmpeg and libmpeg.

I have also looked briefly at freevo, and I am rather tempted on
getting this working as I am a python coder, but it does not have the
feature set that MythTV has.

What do you guys think about starting the cut for 0.22?  Most of the
UI has been ported to MythUI, and it seems like a good time to get
started on stabilising svn-trunk for a release.

I want to work with current MythTV development community to see if we
can quickly find a solution for NZ and Norway.  It is better to role
the support needed into the next MythTV release.  However if there is
not much progress on working together, I will fork 0.21-fixes to get
things done.

(I am an experienced router programmer, C programmer, and have worked
on Linux drivers, and Plone -- lots of experience with patching, patch
editing and using subversion intelligently to manage code
development).

Best Regards,

Matthew grant


More information about the mythtv-dev mailing list