Right now we update QtQuick views after a composition has rendered so
that we'll have them ready for our next frame.
One of my machines (With X11 + nvidia) was quite stuttery, profiling
shows we spend a huge amount of time with the CPU blocked waiting for
the vblank from the previous main compositing to finish swapping before
we could start processing the QtQuick rendering. Time when the CPU is
blocked is wasted time.
This patch does mean kwin's compositing cycle lasts longer but a recent
patch changes the rendering policy when we start a fullscreen effect anyway so
hopefully that should still give us enough time to render that first frame in time.
This change adjusts the window management abstractions in kwin for the
drm backend providing more than just "desktop" outputs.
Besides that, it has other potential benefits - for example, the
Workspace could start managing allocation of the placeholder output by
itself, thus leading to some simplifications in the drm backend. Another
is that it lets us move wayland code from the drm backend.
With fractional scaling integer based logical geometry may not match
device pixels. Once we have a floating point base we can fix that. This
also is
important for our X11 scale override, with a scale of 2 we could
get logical sizes with halves.
We already have all input being floating point, this doubles down on it
for all remaining geometry.
- Outputs remain integer to ensure that any screen on the right remains
aligned.
- Placement also remains integer based for now.
- Repainting is untouched as we always expand outwards
(QRectF::toAdjustedRect().
- Decoration is untouched for now
- Rules are integer in the config, but floating in the adjusting/API
This should also be fine.
At some point we'll add a method to snap to the device pixel
grid. Effectively `round(value * dpr) / dpr` though right now things
mostly work.
This also gets rid of a lot of hacks for QRect right and bottom which
are very
confusing.
Parts to watch out in the port are:
QRectF::contains now includes edges
QRectF::right and bottom are now sane so previous hacks have to be
removed
QRectF(QPoint, QPoint) behaves differently for the same reason
QRectF::center too
In test results some adjusted values which are the result of
QRect.center because using QRectF's center should behave the same to the
user.
ScreenPaintData provides a way to transform the painted screen, e.g.
scale or translate. From API point of view, it's great. It allows
fullscreen effects to transform the workspace in various ways.
On the other hand, such effects end up fighting the default scene
painting algorithm. For example, just have a look at the slide effect!
With fullscreen effects, it's better to leave to them the decision how
the screen should be painted. For example, such approach is taken in
some wayland compositors, e.g. wayfire, and our qtquick effects already
operate in similar fashion.
Given that, strip the ScreenPaintData of all available transforms. The
main motivation behind this change is to improve encapsulation of item
painting code and simplify model-view-projection code in kwin. It will
also make the job of extracting item code for sharing purposes easier.
The Workspace has two stacks - one with managed windows and deleted
windows, the other includes windows from the first stack + override
redirect windows.
This change merges both stacks. It has several benefits - we will be
able to move window elevation stuff to Workspace and streamline the
scene stuff, for example it will be possible to have a root item.
Another advantage is that unmanaged windows will have
Window::stackingOrder() property set, which can be useful in the future
in qml effects or (qtquick scene if we push harder in that front).
Another advantage is that kwin will make less X11 calls when restacking
managed windows.
WindowItem has proper visibility set now, so these two interim hacks can
be removed now.
Workspace::stackingOrderChanged still needs to be kept around as the
Scene has no root item which could track the order of its child (window)
items.
This change makes the WindowItem track the opacity and schedule a
repaint. It further decouples the legacy scene from code window
abstractions.
It's an API breaking change. WindowPaintData no longer can make windows
more opaque. It only provides additional opacity factor.
This allows to toss a large amount of custom rendering code.
Furthermore, it removes the build-time dependency on Plasma Framework
for FrameSvg and Theme from KWin core as it's pulled in through QML
imports now.
It also cleans up the API and removes functions that are effectively
unused or no-op after this change.
For instance, effects often destroy their effect frames
in pre/postPaintScreen, which would now destroy an `OffscreenQuickView`,
which changes GL context. This is alleviated by delaying detruction
of the internal view.
Support for the features of text cross-fade and selection frame,
which are not used by any of the built-in effects, is dropped.
Signed-off-by: Eike Hein <eike.hein@mbition.io>
Since WindowItem::visible is kept in sync with the effective visible
status of the window, window items that are not lockscreen greeter or
input methods can be hidden when the lock state changes.
With this, the WindowItem will know whether it's actually visible. As
the result, if a native wayland window has been minimized, kwin won't
try to schedule a new frame if just a frame callback has been committed.
EffectWindow::enablePainting() and EffectWindow::disablePainting() act
as a stone in the shoe. They have the final say whether the given window
is visible and they are invoked too late in the rendering process.
WindowItem needs to know whether the window is visible in advance,
before compositing starts.
This change replaces EffectWindow::enablePainting() and
EffectWindow::disablePainting() with EffectWindow::refVisible() and
EffectWindow::unrefVisible(). If an effect calls the refVisible()
function, the window will be kept visible regardless of its state. It
should be called when a window is minimized or closed, etc. If an effect
doesn't want to paint a window, it should not call effects->paintWindow().
EffectWindow::refVisible() doesn't replace EffectWindow::refWindow() but
supplements it. refVisible() only ensures that a window will be kept
visible while refWindow() ensures that the window won't be destroyed
until the effect is done with it.
Currently, there's a separate pass to filter out windows not ready for
compositing or windows that must be invisible. That has two issues: we
could merge that pass with the pass that populates stacking_order and
"windows" can detach.
If the window filter rejects a window, that window won't be in the
stacking_order and henceforth won't be painted, so finalDrawWindow()
does extra work of checking again if the window is accepted.
Effects may perform cleanup when a deleted window is removed. If that
happens and the SceneWindow is accessed, kwin may crash.
The Scene processes Workspace::deletedRemoved() before effects.
In order to fix dereferencing null pointer, this change makes the Window
destroy its associated SceneWindow.
This makes KWin switch to in-tree copy of KWaylandServer codebase.
KWaylandServer namespace has been left as is. It will be addressed later
by renaming classes in order to fit in the KWin namespace.
AbstractOutput is not so Abstract and it's common to avoid the word
"Abstract" in class names as it doesn't contribute any new information.
It also significantly reduces the line width in some places.
The main motivation behind this change is to unify render target
representation across opengl and software renderers and avoid accessing
the render backend directory in order to get the render target.
Using the global coordinate system when specifying output layer damage
regions would be very confusing. In order to make the coordinate system
comprehensible, use the layer-local coordinate system.
The infinite region is used to tell the Compositor when it needs to
repaint the entire layer.
The .clang-format file is based on the one in ECM except the following
style options:
- AlwaysBreakBeforeMultilineStrings
- BinPackArguments
- BinPackParameters
- ColumnLimit
- BreakBeforeBraces
- KeepEmptyLinesAtTheStartOfBlocks
Some effects (AnimationEffect) transform windows without setting
PAINT_SCREEN_WITH_TRANSFORMED_WINDOWS flag. Make the scene disable
render optimizations if that's the case.
Whether AnimationEffect does a right thing is up to debate.
Window painting is no longer split in two phases - PAINT_WINDOW_OPAQUE
and PAINT_WINDOW_TRANSLUCENT.
PAINT_WINDOW_TRANSLUCENT is used as a hint to the occlusion culling
logic to ignore the opaque region.
Given that, the handling of the opaque region can be simplified. If no
effect sets the PAINT_WINDOW_TRANSLUCENT flag, then the opaque region
can be used as is.
It's not possible to get the surface damage before calling
Scene::paint(), which is a big problem because it blocks proper surface
damage and buffer damage calculation when walking render layer tree.
This change reworks the scene compositing stages to allow getting the
next surface damage before calling Scene::paint().
The main challenge is that the effects can expand the surface damage. We
have to call prePaintWindow() and prePaintScreen() before actually
starting painting. However, prePaintWindow() is called after starting
rendering.
This change makes Scene call prePaintWindow() and prePaintScreen() so
it's possible to know the surface damage beforehand. Unfortunately, it's
also a breaking change. Some fullscreen effects will have to adapt to
the new Scene paint order. Paint hooks will be invoked in the following
order:
* prePaintScreen() once per frame
* prePaintWindow() once per frame
* paintScreen() can be called multiple times
* paintWindow() can be called as many times as paintScreen()
* postPaintWindow() once per frame
* postPaintScreen() once per frame
After walking the render layer tree, the Compositor will poke the render
backend for the back buffer repair region and combine it with the
surface damage to get the buffer damage, which can be passed to the
render backend (in order to optimize performance with tiled gpus) and
Scene::paint(), which will determine what parts of the scene have to
repainted based on the buffer damage.
We already try to ensure that the surface damage is within render target
bounds. Avoid clipping surface damage in render backend, which is a bit
excessive task and perhaps it should be done an abstraction level above.
If the main surface is translucent (e.g. it contains only the drop
shadow) but its subsurface is opaque, the "window->isOpaque()" check
will produce a false positive.
Software cursor has always been a major source of problems. Hopefully,
porting it to RenderLayer will help us with that.
Note that the cursor layer is currently visible only when using software
cursor, however it will be changed once the Compositor can allocate
a real hardware cursor plane.
Currently, software cursor uses graphics-specific APIs (OpenGL and
QPainter) to paint itself. That will be changed in the future when
rendering parts are extracted from the Scene in a reusable helper.
This is the first tiny step towards the layer-based compositing in kwin.
The RenderLayer represents a layer with some contents. The actual
contents is represented by the RenderLayerDelegate class.
Currently, the RenderLayer is just a simple class responsible for
geometry, and repaints, but it will grow in the future. For example,
render layers need to form a tree.
The next (missing) biggest component in the layer-based compositing are
output layers. When output layers are added, each render layer would
have an output layer assigned to it or have its output layer inherited
from the parent.
The render layer tree wouldn't be affected by changes to the output
layer tree so transition between software and hardware cursors can be
seamless.
The next big milestone will be to try to port some of existing kwin
functionality to the RenderLayer, e.g. software cursor or screen edges.
The responsibilities of the Scene must be reduced to painting only so we
can move forward with the layer-based compositing.
This change moves direct scanout logic from the opengl scene to the base
scene class and the compositor. It makes the opengl scene less
overloaded and allows to share direct scanout logic.
paintScreen() already tries to ensure that the damage region doesn't go
outside the scene geometry. With this change, it will try to clip the
damage region to the render target rect, which saves us an extra region
intersection and simplifies code that calls paintScreen().
