[MPlayer-users] Bicubic scaling in vo_gl - does it ever work?

[Vladimir Mosgalin]

Hi Reimar D?ffinger!

 On 2006.06.24 at 17:13:02 +0200, Reimar D?ffinger wrote next:

> Hello,
> On Tue, Jun 20, 2006 at 04:00:27AM +0400, Vladimir Mosgalin wrote:
> >  On 2006.06.19 at 22:21:16 +0200, Reimar D?ffinger wrote next:
> > > directly to CPU usage: If you enable glfinish you see exactly how much
> > > time the shader uses.
> > 
> > Hm. Well, triple buffering should almost get rid of this effect.. at the
> > price of having shown video frame to lag from rendered one by two
> > frames.
> 
> Not with any conformant OpenGL implementation. glFinish must always wait
> until all pending Operations are completed.

Because of buffering, it still will help if the shader takes different
time to finish depending on current frame. If it's the same, then there
will be no difference.

> Can you try latest SVN? I did some changes that should be unrelated
> though.

It doesn't solve the problem, however now I observe no bugs in 90% of
cases or so, even on the files that almost always failed before.
Changing video controls still bring problems instantly.

On the other hand, it may be about something else, like it started
hapenning after reboot or something.

> The due to the different behaviour for linux/windows it looks like a
> driver bug to me, do you have the latest version?

Yes.. Latest linux driver. Probably not the latest windows one, though..
but that doesn't matter.

New windows and linux drivers are out, but there is nothing useful in
linux driver changelog. I'll update in a few days though.

> If so, will you make a bugreport to ATI? If so, feel free to include my
> email adress for the (unlikely) case that they are seriously interested
> and have questions about the code...

What should I complain about? What exactly testcase should I provide?

> > PS er.. there is no such thing as undefined value in register or
> > something in video card programming, is it?
> 
> Why shouldn't there be such a thing?

Well, don't know. It's just that this problems looks very much like
there is some random value somewhere, and it can either trigger a bug or
it can not. When not using gl for a while, the output will always be
correct on the first run. On the second run, it will most likely be
buggy, but if I wait for a minute, I'll get correct output. When trying
to change b/c/s that value is always non-zero or something due to shader
working at the same moment, so the output gets distorted immediately.

-- 

Vladimir