[MPlayer-users] Do I still load radeon_vid when using vesa:vidix?

Brian J. Murrell 3cc3bad0acf995c5177a1f7dfa673dc5 at interlinx.bc.ca
Sat Feb 9 12:26:02 CET 2002


On Sat, Feb 09, 2002 at 01:30:36PM +0300, Nick Kurshev wrote:
> Hello, Brian!

Hi Nick,

> I suspect that we should disable adaptive deinterlacing for non-interlaced streams

Indeed.  I am playing back MPEG1 streams.  When you quote playing DVD
as being non-juddery, you are playing back MPEG2 streams.  Of course
the former is not interlaced and the latter could be.  Maybe that is
the difference.

> but currently I didn't hack this code in mplayer sources.
> Althrough radeon_vid.so is ready for that ;)

Yeah?  How soon could it be so?  If it's easy enough, it would be nice
to eliminate it as the reason for the judder.

> This question already was answered by me for you in this mlist!

Not in as much detail as you answer here, so I probably didn't
understand.

> vsync is useful only when you use double buffering. (for single buffering
> it's meaningless). When you use -double key in command line then mplayer
> calls FrameSelect function to switch next frame for drawing and
> release the current frame for displaying.

This is an important detail that is not clear or missing in the docs!
I was using -double but only to get rid of the ODS flicker as per the
docs.  It did say there was a CPU penalty for it but I could not
notice any.

> Radeon's driver uses SMART_SWITCH flag of SCALER chip which should be enough
> for this purpose to produce async buffer switching without producing unnecesary
> CPU delays.

Are there any docs available on this operation of the Radeon?

> (Engine SHOULD wait vsync before real updating of internal registers).
          ^^^^^^
Does the use of should here mean that we are all assuming that that is
how it works or is there some documentation somewhere that says that
it works that way?

> That's all.

Cool.  And sounds like it makes sense.  Are we sure it's working?

b.

-- 
Brian J. Murrell




More information about the MPlayer-users mailing list