This is an automatically generated e-mail. To reply, visit: http://git.revie= wboard.kde.org/r/104111/

On February 29th, 2012, 2:26 p.m., Martin G= r=C3=A4=C3=9Flin wrote:

Do I unde=
rstand correctly that it is about this extension: http://www.opengl.org/reg=
istry/specs/SGI/swap_control.txt

If yes I must admit that my brain is today in no condition to understand wh=
at it talks about ;-)

On February 29th, 2012, 2:42 p.m., Thomas L=C3=BCbking wrote:

Yupp, i c=
ontrols the buffer flush interval - like eglSwapInterval is at least suppos=
ed to do. "1" one means "once every frame" while "=
0"  means "whenever you feel like" (values bigger than one w=
ould theoretically be possible, so 4 would flush in a 4 frame interval)

Got a flu?

On February 29th, 2012, 2:49 p.m., Martin Gr=C3=A4=C3=9Flin wrot= e:

ok with 0=
 meaning whenever the driver things it makes sense. And yes I got a flu.

SwapInterva=
l() is unrelated to WaitVideoSync().

It only controls whether SwapBuffers() waits for vblank or not. The default=
 is 1 in all KMS drivers that support syncing to vblank, but the swap behav=
ior can be overridden in a number of places including xorg.conf, ~/.drirc a=
nd by the vblank_mode env var.

SwapInterval() also has no affect on CopySubBuffer()/PostSubBuffer(), but t=
hose functions usually schedule the copy to occur the next time the scanout=
 position is below the rect that's about to be copied, even though they=
 return immediately.

- Fredrik

On February 28th, 2012, 10:08 p.m., Thomas L=C3=BCbking wrote:

Review request for kwin and Martin Gr=C3=A4=C3=9Flin.

By Thomas L=C3=BCbking.

Updated Feb. 28, 2012, 10:08 p.m.

Descripti= on

Believe it or not, but there's actually a glXSwapInterva=
l function and as far as i could figure one should use it since i found con=
tradicting claims about the default (some mesa guy in a mailing list entry =
from 2006 says it's 0, while http://www.opengl.org/registry/specs/SGI/s=
wap_control.txt suggests it's one)

However it can't harm -hopefully- and works here on nvidia and GMA/mesa=
 (nvidia has SGI & ext, mesa has MESA and OML - d'ooohh) and *might=
* be the cause of wonky swapping because if it's not set the color buff=
er flushes whenever it feels like what can cause waitSync to block for even=
 a random amount of frames or not at all (causing tearing) - at least says =
the guy from 2006 and is supported by the reply of the guy who originally a=
sked =3DD

Testing <= /h1>

It's detected and applied, not 100% sure about the resul=
ting impact.

Diffs=

kwin/libkwineffects/kwinglutils_funcs.h (9= 7a7a6c)
kwin/libkwineffects/kwinglutils_funcs.cpp = (6d79507)
kwin/scene_opengl_glx.cpp (ddebcd0)=

View Diff

Descripti= on

Testing <= /h1> It's detected and applied, not 100% sure about the resul= ting impact.

Diffs=

Testing <= /h1>

It's detected and applied, not 100% sure about the resul= ting impact.