A hybrid GPU system is a system where more than one GPU is connected to
connectors. A common configuration is having a integrated GPU (iGPU)
connected to a laptop panel, and a dedicated GPU (dGPU) connected to
one or more external connector (such as HDMI).
This commit adds support for rendering the compositor stage using the
iGPU, then copying the framebuffer content onto a secondary framebuffer
that will be page flipped on the CRTC of the dGPU.
This can work in two different ways: GPU accelerated using Open GL ES
3, or CPU unaccelerated.
When supported, GPU accelerated copying works by exporting the iGPU
onscreen framebuffer as a DMA-BUF, importing it as a texture on a
separate dGPU EGL context, then using glBlitFramebuffer(), blitting it
onto a framebuffer on the dGPU that can then be page flipped on the dGPU
CRTC.
When GPU acceleration is not available, copying works by creating two
dumb buffers, and each frame glReadPixels() from the iGPU EGL render
context directly into the dumb buffer. The dumb buffer is then page
flipped on the dGPU CRTC.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
When creating a renderer with a custom winsys (which is always how
mutter uses cogl) make it possible to pass a user data with the winsys.
Still unused.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Make dumb buffer creation/destruction reusable by introducing a
MetaDumbBuffer type (private to meta-renderer-native.c). This will
later be used for software based fallback paths for copying render GPU
buffers onto secondary GPUs.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Get rid of some technical dept by removing the support in the native
backend for drawing the the whole stage to one large framebuffer.
Previously the only way to disable stage views was to set the
MUTTER_STAGE_VIEWS environment variable to 0; doing that now will cause
the native backend to fail to initialize.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
When drmHandleEvent() returns an error and errno is set to EAGAIN,
instead of ending up in a busy loop, poll() the fd until there is
anything to read.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
The prefix, if any, of a variable name often contains information about
the namespace (such as clutter_backend is the ClutterBackend, while
backend is a MetaBackend). Clean up some more inconsistencies in
meta-renderer-native.c where various variable names were egl_ prefixed
but in fact was Cogl types.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
In order to eventually support multilpe GPUs with their own connectors,
split out related meta data management (i.e. outputs, CRTCs and CRTC
modes) into a new MetaGpu GObject.
The Xrandr backend always assumes there is always only a single "GPU" as
the GPU is abstracted by the X server; only the native backend (aside
from the test backend) will eventually see more than one GPU.
The Xrandr backend still moves some management to MetaGpuXrandr, in
order to behave more similarly to the KMS counterparts.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Move finding, opening and managment of the KMS file descriptor to
MetaMonitorManagerKms. This means that the monitor manager creation can
now fail, both if more than one GPU with connectors is discovered, or
if finding or opening the primary GPU fails.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Turn MetaCrtc into a GObject and move it to a separate file. This
changes the storage format, resulting in changing the API for accessing
MetaCrtcs from using an array, to using a GList.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Turn MetaOutput into a GObject and move it to a separate file. This
changes the storage format, resulting in changing the API for accessing
MetaOutputs from using an array, to using a GList.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
The reverted commit seems to cause
https://bugzilla.gnome.org/show_bug.cgi?id=787240 for some reason. Lets
be safe and revert it for now, as the code freeze is just around the
corner.
This partly (it doesn't reintroduce a whitespace issue) reverts commit
dbc63430d8.
Due to rounding issues, we can't assume a floating point calculation
will end up on an integer, even if we got the factor from the reverse
calculation. Thus, to avoid casting away values like N.999... to N,
when they should really be N+1, round the resulting floating point
calculation before casting it to int.
This fixes an issue where using the scale ~1.739 on a 1920x1080 mode
resulted in error when setting the mode, as the calculated size of the
framebuffer was only 1919x1080.
https://bugzilla.gnome.org/show_bug.cgi?id=786918
When suspending (i.e. VT switching away, the GDM gnome-shell instance
gets hidden, or changing user), destroy the onscreen and offscreen
monitor framebuffers. When resuming, the stage views and framebuffers
will be recreated anyway.
https://bugzilla.gnome.org/show_bug.cgi?id=786299
To support fractional scaling, change the stage view scale to be a
float instead of an int. Also change the places where it is retrieved
and used when scaling things.
https://bugzilla.gnome.org/show_bug.cgi?id=765011
With GLVND, whenever we have both Mesa's and NVIDIA's drives installed
in the system, initializing the GBM backend will always succeed,
regardless of what GPU you have on your system.
This is due to GBM's software rendering fallback.
It seems better to initialize the EGLDevice backend first, which will
fail to find a device match when given a non-NVIDIA GPU.
https://bugzilla.gnome.org/show_bug.cgi?id=784272
Keep track of the logical monitor transform. When a logical monitor is
transformed, all of its monitors are also transformed in the same way.
A logical monitor can either be transformed on the CRTC level, or using
an offscreen intermediate buffer. In both cases will the logical
monitor be transformed, but only in the latter will the view be
transformed.
MetaCrtcs::transform currently does not represent whether the CRTC is
configured to be transformed or not; only when the backend can handle
it does it correctly correspond to the actual CRTC configuration. This
is intended to change with MetaMonitorConfigManager.
https://bugzilla.gnome.org/show_bug.cgi?id=777732
This commit adds support for rendering onto enlarged per logical
monitor framebuffers, using the scaled clutter stage views, for HiDPI
enabled logical monitors.
This works by scaling the mode of the monitors in a logical monitors by
the scale, no longer relying on scaling the window actors and window
geometry for making windows have the correct size on HiDPI monitors.
It is disabled by default, as in automatically created configurations
will still use the old mode. This is partly because Xwayland clients
will not yet work good enough to make it feasible.
To enable, add the 'scale-monitor-framebuffer' keyword to the
org.gnome.mutter.experimental-features gsettings array.
It is still possible to specify the mode via the new D-Bus API, which
has been adapted.
The adaptations to the D-Bus API means the caller need to be aware of
how to position logical monitors on the stage grid. This depends on the
'layout-mode' property that is used (see the DisplayConfig D-Bus
documentation).
https://bugzilla.gnome.org/show_bug.cgi?id=777732
Expose via a new API whether the transform on a logical monitor is
handled by the backend. This was previously only exposed only in the
native backend. This will be used to emulate not supporting transforms
in the backend in the nested backend.
https://bugzilla.gnome.org/show_bug.cgi?id=779745
Whenever an EGLOutput consumer is temporary unable to handle
eglStreamConsumerAcquire() operations (e.g. during a VT-switch),
an EGL_RESOURCE_BUSY_EXT error is generated.
This change adds the appropriate error handling to flip_egl_stream() in
order to recover from such errors.
https://bugzilla.gnome.org/show_bug.cgi?id=779112
In preparation for further refactorizations, rename the MetaMonitorInfo
struct to MetaLogicalMonitor. Eventually, part of MetaLogicalMonitor
will be split into a MetaMonitor type.
https://bugzilla.gnome.org/show_bug.cgi?id=777732
We need to do swap notifications asynchronously from flip events since
these might be processed during swap buffers if we are waiting for the
previous frame's flip to continue with the current.
This means that we might have more than one swap notification queued
to be delivered when the idle handler runs. In that case we must
deliver all notifications for which we've already seen a flip event.
Failing to do so means that if a new frame, that only swaps buffers on
such a swap notification backlogged Onscreen, is started, when later
we get its flip event, we'd notify only an old frame which would hit
this MetaStageNative's frame_cb() early exit:
if (global_frame_counter <= presented_frame_counter)
return;
and we'd never finish the new frame and thus clutter's master clock
would be waiting forever stuck.
https://bugzilla.gnome.org/show_bug.cgi?id=774557
When flush-swap-notify is already queued, we might end up trying to
requeue it, for example when handling flip callbacks inside
swap-buffers. Actually queuing it there is harmless, since old frames
will be discarded anyway.
https://bugzilla.gnome.org/show_bug.cgi?id=774923
We might still end up in swap-buffer without the previous flip callback
having been invoked. This can happen if there are two monitors, and we
manage to draw before having all monitor flip callbacks invoked.
https://bugzilla.gnome.org/show_bug.cgi?id=774923
This commit adds for a new type of buffer being attached to a Wayland
surface: buffers from an EGLStream. These buffers behave very
differently from regular Wayland buffers; instead of each buffer
reperesenting an actual frame, the same buffer is attached over and
over again, and EGL API is used to switch the content of the OpenGL
texture associated with the buffer attached. It more or less
side-tracks the Wayland buffer handling.
It is implemented by creating a MetaWaylandEglStream object, dealing
with the EGLStream state. The lifetime of the MetaWaylandEglStream is
tied to the texture object (CoglTexture), which is referenced-counted
and owned by both the actors and the MetaWaylandBuffer.
When the buffer is reattached and committed, the EGLStream is triggered
to switch the content of the associated texture to the new content.
This means that one cannot keep old texture content around without
copying, so any feature relying on that will effectively be broken.
https://bugzilla.gnome.org/show_bug.cgi?id=773629
This commit adds support for using a EGLDevice and EGLStreams for
rendering on top of KMS instead of gbm. It is disabled by default; to
enable it pass --enable-egl-device to configure.
By default gbm is first tried, and if it fails, the EGLDevice path is
tried. If both fails, mutter will terminate just as before.
https://bugzilla.gnome.org/show_bug.cgi?id=773629
There is no way to pass any backend specific parameters to a
CoglFramebuffer until after it has been allocated by
cogl_framebuffer_allocate() (since this is where the winsys/platform
fields are initialized). This can make it hard to actually allocate
anything, if the platform depends on some backend specific data.
A proper solution would be to refactor the onscreens and framebuffers to
use a GObject based type system instead of the home baked Cogl one, but
that'll be left for another day. For now, allocate in two steps, one to
allocate the backend specific parts (MetaOnscreenNative), and one to
allocate the actual onscreen framebuffer (via
meta_onscreen_native_allocate()).
So far there is nothing that forces this separation, but in the future
there will, for example EGLDevice's need to know about the CRTC in
order to create the EGLSurface.
https://bugzilla.gnome.org/show_bug.cgi?id=773629
A swap-buffers should never be issued when we are waiting for a flipped
callback, so instead of trying to handle a situation that sholud never
happen, warn instead.
https://bugzilla.gnome.org/show_bug.cgi?id=773629
When a swap failed with EACCES (possibly due to VT switching), don't
mark the framebuffer as 'in use', so that it'll be cleaned up properly
and not set as current.
https://bugzilla.gnome.org/show_bug.cgi?id=773629
Drivers may be bad at guessing what is passed to eglGetDisplay, ending
up return non-functioning EGLDisplay's. Using eglGetPlatformDisplay
avoids this issue.
https://bugzilla.gnome.org/show_bug.cgi?id=773629
