[mythtv-users] he shoots. he scores!

Coax coax at cornernet.com
Thu Apr 1 17:40:56 EST 2004


Interesting, Matthias.  Thank you for your input!

I really do wonder.. I haven't tested with the ivtvdev driver since I got
my Myth box up and running.  I do have the driver installed and have
tested it.. just never saw a reason to use it, as i didn't notice a
difference in performance. (Duh, i'm out'ing to /dev/video16, so it
wouldnt really matter! heh :)  I'm going to have to try this, given your
response.

The question, however, does remain:  Is there a way, within MythTV, to
support this sort of thing natively?

I.e. here's an example:

If I happen to have a PVR-350 and use it for TV-Out and capture, and I
*ALSO* happen to have a BT878 dumb framegrabber in the box..  MythTV is
going to capture on /dev/video0 and write the frames it got as .nuv files
to disk.  However, the dumb framegrabber is going to capture raw frames,
be software encoded as MPEG-4.

I want the MPEG-II encoded files to "co-exist" with the MPEG-4 encoded
files, which I believe MythTV already does natively. (And works well, when
software decoding and displaying on a normal video card.)

I wonder how to make MythTV know that it can't play MPEG-4 files through
the /dev/video16 device on the PVR-350, so not to bother - and rather, to
play them via the normal X display instead..

Thoughts, anyone?

Chad

P.S. I kinda-sorta asked a question on these lines a few days ago.. I got
one response which fully answered my question, however, would still like
to be able to use my PVR-350 and dumb capture card in the same box without
having to buy a different video card for TV-Out.

 > I do agree but one small detail though.
> You can access the OSD buffer (what you call framebuffer) two ways:
> - mmap directly the ram of the card and writing in it => slow
> - dma the buffer you want => fast.
> You get a confortable 25fps display with divx/dvd.
> There IS some artifacts though as said below but it is better than nothing.
>
> One last detail : if you use the normal X driver, it does the mmap stuff.
> Use the driver from http://membres.lycos.fr/badzzzz which is using the DMA technique.
>
> We are not there quite yet but getting there.
> M
>
>
> Niklas Brunlid <nbr at ticalc.org> writes:
>
> > On Tue, 30 Mar 2004 17:13:55 -0600 (CST), Coax <coax at cornernet.com> wrote:
> >
> >> Hello, Ivan.
> >>
> >> There is 1 good reason to not use the TV-Out on the PVR-350 for all
> >> MythTV
> >> applications.
> >>
> >> Its framebuffer (i.e. that which is NOT displayed on /dev/video16) is so
> >> painfully slow, it rivals the first Trident PCI video card ever made. :)
> >> (It doesn't even seem to support XvMC extensions!)
> >>
> >> This means it is painfully slow to play anything that is NOT encoded with
> >> MPEG-2..
> >>
> >> I have a PVR-350 in my MythTV box (athlon "2200+", which is the 1.8ghz
> >> 266fsb chip...) and if I just so happen to disable hardware decoding -
> >> and
> >> just out to the framebuffer directly, its like watching TV in
> >> slow-motion.
> >> (Meanwhile, the box is absolutely flogged.  load skyrockets.)
> >
> > Why do people keep saying this? I can guarantee that the hardware
> > framebuffer on the PVR350 is plenty fast enough to play video in
> > fullscreen, full frame rate. On my system (Pundit, P42.4GHz, 512MB
> > somethingSpeed memory, MythTV 0.14, Gentoo gentoo-dev-sources 2.6.4-r1,
> > Mathias' patched and ready-to-use 0.1.9a ivtv drivers) it simply works,
> > and uses ca 20% CPU for mplayer software-scaled output, less for DVD with
> > Xine.
> >
> > As stated many times, once you have the speed, the problem becomes
> > tearing, i.e. the players output is not synced with the PVR350:s output.
> > Especially noticable on panning scenes or high-action scenes, but
> > otherwise fine.
> >
> > I honestly haven't tried using software decoding for PVR350-recorded
> > content (at least not in MythTV) since I haven't had a reason. =)
> >
> > Also, Mathias mentioned a patch that encodes video output to MPEG2. Could
> > be interesting, haven't tried it myself though.
>



More information about the mythtv-users mailing list