[mythtv] Re: [mythtv-commits] Bob and XVMC
Ivor Hewitt
ivor at ivor.org
Sat Feb 5 14:27:00 UTC 2005
On Friday 04 Feb 2005 23:55, Daniel Thor Kristjansson wrote:
> On Fri, 4 Feb 2005, Ivor Hewitt wrote:
> ]Yup indeed fixes XvMC. Green flickering all over screen with just XV.
>
> Are you getting the flickering in XV with the current CVS as of this
> afternoon?
>
Hi Daniel,
Yup XV still flickering and I can't work out what's going on. I can remove the
mad green flashing by removing the memcpy's in PrepareFrame for GUID_YV12..
but I don't understand what that's supposed to be doing...
Then with that removed the behaviour is that the first field is displayed
fine, but the second field is displayed with inverted colours. Hence some
rather strange flickering effects. I thought it was just a memory alignment
problem, but it doesnt appear to be.... I guess it's a bug somewhere in the
unichrome driver.
> ]Out of interest what is the halfLineSrc adjustment supposed to be doing
> for ]XvMC? I thought the field 1 and 2 values were what told the XvMC
> driver which ]field was active. So why does the image need offsetting?
>
> No idea why it's need, I think the driver should take care of it. It may
> be just an nvidia driver bug. But XvMC standard is not clear on who
> should do the offsetting, so doing or not doing the offsetting may both
> be proper for the driver writer.
In that case I don't understand the bob logic for the PutImage calls. The
adjustments to the y position affect the destination position not the source.
The logic appears to only make sense where dest resolution is the same as the
source resolution, I can't see how it could/would work otherwise....
For example say the source fields are vertically interleaved frames A and B:-
ABABAB
which is scaled (badly) to the destination say to:-
ABBABBAB
Offsetting the destination will result in alternating between:
ABBABBAB and
ABBABBAB
Which will look orrible. Surely just the source position needs offsetting to
produce alternating:-
ABBABBAB and
BAABAABA
Regards,
Ivor.
More information about the mythtv-dev
mailing list