[mythtv-users] fps drops with DRM sync and bob

Doug Larrick doug at parkercat.org
Wed Jun 6 21:49:48 UTC 2007


Hamish Moffatt wrote:
> I've tweaked my modeline to get ~50.002Hz, but 50.0 simply isn't
> possible. I cut up myth's vsync.cpp to write a simple test program to
> measure the DRI vblank interval, and it's reporting 50.002Hz.
> 
> When I start playing video with bob (or kerneldeint) there's not much
> change in the result, though it does jitter a bit more than when just
> sitting in the static MythTV menus.
> 
> However bob still goes nuts after a few seconds; dropping frames to
> catch up with video. This is with audio as timebase. The picture is
> rather jerky as frames are dropped.

This is odd.  From what you describe you should be fine.  I have a
similar refresh disparity on my dev system (60.0 Hz vs. 59.97 Hz) and
all I see is an occasional doubled frame.

> I notice this seems to occur with video as timebase as well, but in that
> cause shouldn't audio frames be dropped rather than video?

You would think.  I've never had much success with video as timebase
(even though it should, in theory, be superior -- this is what Tivo uses
for example).

> Maybe I need to measure my audio output rate too, though I don't see wy
> bob and kerneldeint would differ in behaviour if audio was the cause of
> the problem. My system has Intel HDA audio on board.

Hmm, my dev system is also Intel graphics (965) and Intel HDA audio; I
run bob without a problem.  What kernel are you running? (I am at
2.6.21.1 currently.)  Intel HD Audio is new enough that maybe there are
bugs in older drivers.

> Before it goes nuts, the picture shimmers quite a bit vertically. 
> Also I notice a lot of stretched frames and things when seeking which I
> never saw on the old NVIDIA system.

Is it possible the progressive vs. interlaced detection is going wrong?
 Can you try forcing one or the other from the OSD menu?

The only artifact I get on my dev system is tearing on HD-sized frames;
this is a known current limitation of the driver.

> Is the DRI sync well tested? If NVIDIA doesn't support it I wonder how
> many people are actually using it.

Yeah, like I said it's working fine for me on my dev system (which
granted is only one datapoint but still).  DRM vsync is also used by ATI
folks.

The most recent Intel xorg drivers (I'm sure new Mesa is required as
well) do support the OpenGL vsync extension (at least on 965), so you
could rule out DRM sync problems that way.

-Doug

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 252 bytes
Desc: OpenPGP digital signature
Url : http://mythtv.org/pipermail/mythtv-users/attachments/20070606/26f11bf0/attachment.pgp 


More information about the mythtv-users mailing list