[mythtv] Re: Mythtv HMC/Development 1 - Digital and High Definition
Daniel Thor Kristjansson
danielk at cat.nyu.edu
Tue Mar 30 12:59:52 EST 2004
My patches are generally available here:
Unless I've recently submitted them for consideration in CVS.
These patches are a little over a month old at the moment. Until
last evening I was working on a big project at my day job. I will be
updating those sometime in the next few days.
For HDTV vs. SD the problem is that decoders have a really hard
time keeping up at a high resolution. Some have thrown fast processors
at the problem and have had success with software decoding. I would like
for this to work with some kind of hardware assistance and a less than
state of the art processor. One problem with this is that XvMC on most
cards does OSD with a full copy of the frame, which is very slow.
Another is that, at least with nvidia, you only get eight XvMC buffers
total. This isn't enough to cope with fast scenes. Finally there is a
problem with ffmpeg in that it doesn't allow for a gradual transition
in frame skipping, it's either all frames or no b-frames at all.
For TS streams in general ffmpeg is not the most mature decoder. It
throws away redundant TS packets, even if the PES packet fails the
CRC check. It also does not deal with bad streams with grace, if the
CRC by chance passes on a bad PES packet you will likely dump core
Problems that are more Myth related on the backend side come with
decoding ATSC info, which you won't have to deal with on the DVB
side. And tuning, which you will have to deal with. I patched Myth to
add a signal check to HDTV tuning, I tried a blind port to the DVB
side, but I didn't get any reports of success with it. This would
make recovery from tuning to an encrypted channel more sane, and with
a few tweaks, faster.
Oh, another problem is with switching between resolutions, this
sometimes results in a massive spew of errors about invalid blocks.
Sometimes this crashes Myth. This is just a bug in MythTV, it needs to
pay attention to changes in the resolution of the stream and wait for
I-frames before decoding anything after a resolution change (which may
require ffmpeg cooperation). But it needs to be tackled at some point.
Deinterlacing HDTV is also an issue, but I think that will be solved
relatively soon so you shouldn't worry too much about it.
BTW The hdtv-playback patch does frame skipping, but it's the all
b-frame type, and the heuristic for turning frame skipping on and off
is not very clever.
On Tue, 30 Mar 2004, Nick Croucher wrote:
] I have been given the task of getting the dvb support working in
]the HMC (development one?s) derivative of Mythtv. I have only just started
]looking into the code for this.
]At this point I can successfully view all the free digital stations (SD and
]HD) here in Melbourne using tzap and mplayer. I get reasonable (but not
]quite stable) results with Mythtv 0.14 watching SD channels but get a lot of
]errors when I try HD channels. I have just built a cvs Mythtv, but if I try
]to view any dvb channel or recording in kills the front end :-(.
]The cvs version is what I?m most interested in and will start to try and fix
]it up a bit. I am willing to share all of my findings and fixes (if any :P )
]with you or with the forums directly, but as you may be able to appreciate,
]our current version of Mythtv differs a bit from what you guys have. That
]being said I would like to see the cvs version working well and then I will
]back port the relevant stuff into our version.
]I haven?t looked at your HD patches yet, but will do so tomorrow (if you
]could email them to me that would be real nice), and I will post on the
]developers list after I have look at the code a bit (probably tomorrow as
]Well I look forward to collaborating with you guys, hopefully you can all
]put up with a hack like me :P
]Outgoing mail is certified Virus Free.
]Checked by AVG anti-virus system (http://www.grisoft.com).
]Version: 6.0.605 / Virus Database: 385 - Release Date: 1/03/2004
More information about the mythtv-dev