[mythtv-users] Advanced 2X not working after system upgrade

[Neil Salstrom]

On Tue, Oct 2, 2012 at 6:35 PM, Justin Johnson
<justin.johnson3 at gmail.com> wrote:
> Hi everyone,
> I've upgraded from Mythbuntu 10.04.3 to 12.04.1 this weekend and am no
> longer able to successfully use the VDPAU advanced 2x deinterlacer.
> I am using the techniques outlined in the Judder free wiki page to
> synchronize refresh rates, my TV's supported modes are:
> *1920x1080 at 60
> *1920x1080 at 24
> *1920x1080 at 23.976
> *1920x1080 at 60i
> *1920x1080 at 59.94
> *1920x1080 at 59.94i
> *1280x720 at 60
>
> In Xorg.0.log I see the screen change to 59.94Hz when I start
> playback, and can confirm with the info display on my TV and xrandr.
> However, the Advanced2X deinterlacer is not working. I can see the
> interlacing artifacts on the screen, and have seen at least once in
> the playback log something along the lines of "monitor does not
> support double framerate". The Advanced1X deinterlacer does work,
> though it doesn't appear to be perfectly smooth.
>
> Important version number:
> NVIDIA driver: 304.51
> Linux kernel 3.2.0-31-generic
> Xorg: 1.11.3
> Mythtv:  v0.25.2-31-g33c34da; compiled myself, with the only change
> from fixes/0.25 being
> http://code.mythtv.org/trac/changeset/21306c39/mythtv
>
> I posted my full xorg.conf at http://pastebin.com/yM3smYbP
>
> Any help troubleshooting would be very much appreciated.
>
> --
> Justin Johnson
> _______________________________________________
> mythtv-users mailing list
> mythtv-users at mythtv.org
> http://www.mythtv.org/mailman/listinfo/mythtv-users

The 3XX.XX series of Nvidia drivers handles modelines totally
differently.  I've not been able to get it working well myself so I'm
on 295.75.  I either lose modes or HD-audio (long story there).

Here is a quote from an email chain I've had going on with an Nvidia developer:

"There was a significant change to the mode validation process in the
302.* driver series: previously, a "mode" consisted of a set of
"back-end timings," i.e. the timings the GPU actually sent to the
monitor, and "front-end timings," which were reported to the X server.
 The front-end timings were basically only used for their height and
width, with the rest of the parameters ignored.  That lead to all
kinds of confusion when, for example, programs would calculate and
report the "refresh rate" of the front-end timings, even when that was
a lie and the back-end timings were completely different.

302.* did away with all that.  Now, the modes reported by the driver
are the real timings sent to the monitor.  This means that only modes
that your monitor can support are advertised.  In particular, if your
flat panel reports that only its native mode is supported, then only
its native mode will appear in the mode pool.  To get scaling, you can
use the new ViewPortIn and ViewPortOut attributes on MetaModes,
documented here:
ftp://download.nvidia.com/XFree86/Linux-x86/304.37/README/configtwinview.html

You can also get the driver to automatically include MetaModes that
scale to common resolutions via the IncludeImplicitMetaModes option
(which is on by default).  These MetaModes are reported through the
RandR 1.1 mode list: you can query them with "xrandr --q1" and set
them with "xrandr -s <size number> -r <fake refresh rate number>".
They are *not* reported through RandR 1.2, since 1.2 gives
applications complete control over the display configuration.

I know it's confusing.  We're working on making nvidia-settings better
at coping with the new configuration flexibility.  In the meantime,
let me know if you have any questions about how to configure your
displays. "