This reduces the repaints in cases where several blur regions are
stacked on top of each other (e.g. oxygen-transparent) and the topmost
layer needs to be updated (e.g. a blurry window is moved).
The old blur version wrongly marked a cache region as valid with the
reason that this region would never become visible. It didnt matter
because the only case that this region could become visible was a
movement of the window on top, which back then would have forced a
workspace repaint of that region and as such would have invalidated
the cache anyway. With the introduction of addLayerRepaint the
last point is no longer true and we have to track the valid cache
regions more carefully.
This patch adds a new function Toplevel::addLayerRepaint, that in contrast
to addWorkspaceRepaint does not invalidate every blur texture cache that
overlaps with that region. As the name suggests it rather invalidates the
to the window associated layer at that position. This is especially useful
in the case of move/resize events in combination with oxygen-transparent,
where the altered window is almost always the topmost window and the blur
texture cache of the windows underneath are unchanged.
For the case of fully opaque windows the behaviour of addLayerRepaint
and addWorkspaceRepaint should be same.
REVIEW: 103906
Too many crash reports in the fglrx driver when using ARB shaders.
It's a pity to disable functionality for all fglrx users, but it's
better than crash reports.
Blur effect can still be enabled manually and Lanczos can still be
enabled through the environment variable.
REVIEW: 103556
CCBUG: 270818
CCBUG: 286795
Finally it is possible :-) If it causes an impact on performance
during the release candidates I'm going to revert.
REVIEW: 103375
BUG: 255106
FIXED-IN: 4.8.0
This fixes graphical glitches that appear next to the borders of plasma
tooltips. Additionally I added a variable that forces the effect to
drop the cache, which might be useful in the future to fix bugs.
This should fix 2 issues with the blur effect:
- It is now sufficient to have a damaged EXPANDED blurArea to trigger
an update of the caching texture.
- No glitches in cases where the taskbarthumbnail effect paints a
thumbnail of a blurred window.
CCBUG: 288568
This patch introduces some kind of damage propagation. In the old version of the blur effect
we had to repaint the whole window if sth behind the blurred area was damaged. The new texture
cache, which was introduced by the last patch, gives us the opportunity to only update parts of
blurred background texture. This means that the damaged area can only propagate with the speed
of the blurring radius per window layer.
REVIEW: 102665
This patch adds an optional texture cache to the blur effect such that damaged windows in
front of the blurred region dont trigger a repaint of the whole blurred region which pretty
often results in a avalanche repaint of nearly the whole screen.
REVIEW: 101977
This is actually a bloody hack by forcing the screen to be
repainted in each frame. This needs to be fixed in a different
way or we at least have to turn blur off by default for GLES in
the next release.
Optimize the region of the screen that's painted in each frame,
and paint the windows top-to-bottom when the damaged region doesn't
intersect any blurred areas.
All previously existing windowAdded methods are renamed to slotWindowAdded.
EffectsHandlerImpl is connected to Workspace's clientAdded signal, which is
emitted a little bit earlier than the previous direct method call. This might
change behavior.
Another signal is added to Workspace to signal that an unmanaged is added.
Since the EffectFrames have been moved into KWin core nothing in the
Effects lib actually used Plasma. The only remaining method is moved
to core as it's not used in the Effects. The Effects itself still
link against Plasma, so nothing changes for them.
The Plasma includes in the kwineffects header seemed to pull in
quite some additional headers, so the includes in some effects have
to be adjusted (most often KConfigGroup). This should speed up the
compilation of the library and the effects.
It now uses a GLShader for GLSL shaders and pushes it using the
ShaderManager.
It does not work with the nouveau driver plus GLES, but it works
with fglrx + desktop GL 2.x, so I assume it is a driver problem here.
Adding a new AbilityUsesBlurBehind which can be set by decorations with
alpha channel to request that the background can be blurred. This improves
the situation for our completely opaque default Oxygen decoration.
This change is a behavior change in comparison to 4.5! Translucent decorations do
not get blurred by default any more.
CCMAIL: craig.p.drummond@googlemail.com
svn path=/trunk/KDE/kdebase/workspace/; revision=1195274
BlurEffect::supported().
- Don't announce support to clients if the shader failed to compile,
or the FBO is invalid.
svn path=/trunk/KDE/kdebase/workspace/; revision=1179268
Currently only scaled or translated windows were excluded,
but it is possible to also have windows with just changed
window quads (e.g. magic lamp).
CCBUG: 243693
svn path=/trunk/KDE/kdebase/workspace/; revision=1169160
It uses a new config group Blacklist in kwinrc with two subgroups for Lanczos and Blur.
The key of each entry in the group referrences the vendor either using GL_VENDOR or GL_RENDERER, e.g. "Intel" in renderer for Intel based chips.
The value of the entries is a stringlist containing exactly the renderer information. This way it is possible to blacklist several mesa driver versions which are buggy.
The blacklist is currently empty and will be filled with a kconf update script (which could also be done by distributions). So all users who experience problems and are subscribed to the bugs receiving this message, please upload your glxinfo.
An example for a blacklist:
[Blacklist][Lanczos]
NVIDIA=GeForce 9400M/PCI/SSE2
CCBUG: 243181
CCBUG: 242985
svn path=/trunk/KDE/kdebase/workspace/; revision=1146195
Used in cube animation for all sticky windows and in desktop grid for the add/remove desktops window.
svn path=/trunk/KDE/kdebase/workspace/; revision=1137489
The texture needs to be drawn on screen coordinates on the offscreen
framebuffer in the horizontal pass, otherwise pixels that fall below
the screen won't be clamped to the screen edge in the subsequent
vertical pass. This is because the offscreen buffer is the same size
as the screen.
This change also gets rid of the need to clear the offscreen buffer
before blurring the screen contents behind each window, and reduces
the CPU computations done on the blur region.
svn path=/trunk/KDE/kdebase/workspace/; revision=1137134
Martin Gräßlin
Philipp Knechtges
Thomas Lübking
Fredrik Höglund
Martin Gräßlin
Kevin Ottens
Andreas Demmer