[mythtv] Re: [mythtv-commits] mythtv commits

Bruce Markey bjm at lvcm.com
Wed May 4 23:15:24 UTC 2005


Daniel Kristjansson wrote:
> On Wed, 2005-05-04 at 14:17 -0400, Daniel Kristjansson wrote:
> 
> 
>>BTW I'm working on a bob fix that doesn't break XvMC, I'll probably send
>>a patch tonight if my testing goes well.
> 
> 
> Attached is a patch that fixes bob displacement for overscan
> with XVideo, while letting XvMC still work when there is no
> overscan.
> 
> It appears the nVidia closed source drivers no longer
> require a bob displacement, as of v66.29. Of course, this
> is a blessing and a curse, the automatic displacement
> means we don't need a displacement normally, but it does
> not work at all when we do over/under scan.
> 
> I've tried the obvious of applying displacement only in the
> overscan or underscan cases, and also the less obvious of
> calculating the normal displacement and over/under scan
> displacement then applying the difference. Neither worked.
> 
> Anyway, I'd like to know if the patch is a regression
> for anyone before applying it.

Using the same Xv systems as before, TVout and SVGA fullscreen
and windowed all with bob turned on. This seemed to work correctly
in almost every case. Using vertical overscan at 0, +6, -6, +20
and -20, the images were all centered correctly and, with one
exception, the two fields meshed correctly. In all cases, when
pressing Pause and unpause the still picture aligned with the
moving picture down to the exact scanline (I don't believe this
has always been true in the past).

The one anomaly was the TVout from a "GeForce4 MX 420" was correct
for all negative values and positive values up to 3%. At four
percent or higher, the fields were off by one scan line causing
doubled horizontal edges. This was not the case for TVout from
a 5200 or monitor output so I believe this was an anomaly of
that card/driver rather than a miscalculation.

Here is the output from a 1400x1050 monitor. 640x480 GUI with
fullscreen playback and vertical overscan at +6% . The values
appear to be correct including xv_dest_y_incr=2 to account for
the scaling.

Total desktop width=1400, height=1050, numscreens=1
2005-05-04 15:53:10.069 Running in a window
2005-05-04 15:53:10.069 Using screen 0, 1400x1050 at 0,0
[...]
2005-05-04 15:53:19.502 Using realtime priority.
2005-05-04 15:53:19.577 Changing from None to WatchingLiveTV
2005-05-04 15:53:19.674 Video timing method: SGI OpenGL
####### Field 1 #######
         src_y: 28
        dest_y: 0
 xv_src_y_incr: -14
xv_dest_y_incr: 0
      disphoff: 1050
          imgh: 422

####### Field 2 #######
         src_y: 28
        dest_y: 0
 xv_src_y_incr: 226
xv_dest_y_incr: 2
      disphoff: 1050
          imgh: 422

--  bjm



More information about the mythtv-dev mailing list