Instead of change window attributes back and forth, calculate the evmask
and set it just once. And also make sure the request is flushed.
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
There is actually no specification what symbols are exported from a
libGL implementation. The (extremely outdated) OpenGL ABI specification
says only GL 1.2 functions are guaranteed. Don't know how relevant that
is now, but different libGL implementations do export different set of
symbols. On Linux we are most likely to be linked with libglvnd, which
has everything we need. But on other platforms this is not necessarily
the case, for example on OpenBSD we are missing glGetQueryObjectui64v.
Use libepoxy so we can outsource this problem and never worry about it
ever again. Plus it also saves us from calling GetProcAddress ourselves.
Changes other than trivial build fixes I have to make:
1. Can't use eglCreatePlatformWindowSurface/eglGetPlatformDisplay.
libepoxy checks for EGL 1.5 when resolving these functions. But
without a current context, libepoxy assumes we only have EGL 1.4.
This creates a chicken and egg problem - we need a display to call
eglGetPlatformDisplay. We have to use the *EXT version instead.
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
I think I was trying to avoid introducing pthread as a dependency, but
now we are using pthread for SGI_video_sync thread anyway. Let's remove
the ifdefs.
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
OpenBSD don't have support for sched_getparam(), sched_setparam(), or
sched_setscheduler() functions (yet). In this case, we need use
pthead-equivalents for real-time sched for picom. Theses changes add
this support.
Authored-by: Jose Maldonado aka Yukiteru <josemald89@gmail.com>
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
Image decouple should keep all the image properies from the source
image, so shader must be copied. And there are also some internal
properties what should be inherited but wasn't.
In particular this prevents images from losing their shaders when
alpha is applied.
Fixes#1174
fix the same issue in the legacy backends, see the previous commit for
details. this commit also removes the x_validate_pixmap function because
it was used only in the get_root_tile function and the fix pretty much
implies and embeds it.
if the root background pixmap's depth doesn't match the root window's
depth, find a suitable visual for the root background pixmap's depth and
use it instead of the root window's visual
This is a follow up to 81d137a3cc and
bug #704. Basically a window will have a `XCB_NONE` as `client_win` if its
previous client_win detached and then the window itself is immediately
destroyed. Because the window is destroyed we couldn't call
`win_recheck_client` so its `client_win` will remain `XCB_NONE`.
However, it turns out we have a convention of setting `client_win` to
the window itself if windows that don't have a client window. So make
sure this convention is followed even for destroyed windows.
Doesn't really fix anything, just to make sure an invariant holds.
Related: #704
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
Present extension based scheduler doesn't work well on NVIDIA drivers.
GLX_SGI_video_sync is less accurate, but is rumoured to work. See
[kwin's usage](https://invent.kde.org/plasma/kwin/-/blob/master/src/
backends/x11/standalone/x11_standalone_sgivideosyncvsyncmonitor.cpp)
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
It's kind of dumb anyway. If we get damage event right after a vblank
event, we would waste the whole vblank.
Instead improve the frame scheduling logic to target the right vblank
interval. This only affects smart_frame_pacing anyway.
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
Right now we rely on `reschedule_render_at_vblank` being scheduled for
the render finish check. Which means we will only know if a frame has
finished rendering the next time we queue a render. And we only check at
vblank boundary, which means when a render is queued we have to wait for
the next vblank, when had we checked earlier, we will be able to start
rendering immediately
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
I noticed sometimes full frame rate video is rendered at half frame rate
sometimes. That is because the damage notify is sent very close to
vblank, and since we request vblank events when we get the damage, we
miss the vblank event immediately after, despite the damage happening
before the vblank.
request next ...... next next
damage vblank vblank vblank
| | | ...... |
v v v v
---------------------->>>>>>---------
`request vblank` is triggered by `damage`, but because it's too close to
`next vblank`, that vblank is missed despite we requested before it
happening, and we only get `next next vblank`. The result is we will
drop every other frame.
The solution in this commit is that we will keep requesting vblank
events, right after we received a vblank event, even when nobody is
asking for them. We would do that for a set number of vblanks before
stopping (currently 4).
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
present_notify_msc with divisor == 0 has undocumented special meaning,
it means async present notify, which means we could receive MSC
notifications from the past.
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
Also, vblank event callback should call schedule_render to queue renders
instead of starting the draw timer directly, so that the CPU time
calculation will be correct.
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
Disable timing estimation based pacing by default, as it might not work
well across drivers, and might have subtle bugs.
You can try setting `PICOM_DEBUG=smart_frame_pacing` if you want to try
it out.
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
Factored out vblank event generation, add abstraction over how vblank
events are generated. The goal is so we can request vblank events in
different ways based on the driver we are running on.
Tried to simplify the frame scheduling logic, we will see if I succeeded
or not.
Also, the logic to exclude vblank events for vblank interval estimation
when the screen off is dropped. It's too hard to get right, we need to
find something robust.
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
