Turns out inverting a matrix was the largest chunk of the CoglMatrix
code. By switching to Graphene, a lot of it can go away. The inverse
is still cached in the CoglMatrix struct itself, to preserve the
optimization.
However, switching to graphene_matrix_t to calculate the inverse has
a challenge: float precision. We had to work around it here, and it
needs an explanation.
The way to detect whether a matrix is invertible or not (i.e.
whether it's not a "singular" matrix, or not) is by checking
if the determinant equals 0. So far, so good.
Both graphene_matrix_t and CoglMatrix use single-precision
floats to store their 4x4 matrices. Graphene uses vectorized
operations to optimize determinant calculation, while Cogl
tries to keep track of the matrix type and has special-purpose
determinant functions for different matrix types (the most
common one being a 3D matrix).
Cogl, however, has a fundamentally flawed check for whether
the matrix is invertible or not. Have a look:
```
float det;
…
if (det*det < 1e-25)
return FALSE;
```
Notice that 1e-25 is *way* smaller than FLT_EPSILON. This
check is fundamentally flawed.
"In practice, what does it break?", the reader might ask.
Well, in this case, the answer is opposite of that: Cogl
inverts matrices that should not be invertible. Let's see
an example: the model-view-projection of a 4K monitor. It
looks like this:
```
| +0,002693 +0,000000 +0,000000 +0,000000 |
| +0,000000 -0,002693 +0,000000 +0,000000 |
| +0,000000 +0,000000 +0,002693 +0,000000 |
| -5,169809 +2,908017 -5,036834 +1,000000 |
```
The determinant of this matrix is -0.000000019530306557.
It evidently is smaller than FLT_EPSILON. In this situation,
Cogl would happily calculate the inverse matrix, whereas
Graphene (correctly) bails out and thinks it's a singular
matrix.
This commit works around that by exploiting the maths around
it. The basis of it is:
inverse(scalar * M) = (1/scalar) * M'
which can be extrapolated to:
inverse(M) = scalar * inverse(scalar * M) = M'
In other words, scaling the to-be-inversed matrix, then
scaling the inverse matrix by the same factor, gives us
the desired inverse. In this commit, the scale is calculated
as 1 / (smallest value in the diagonal).
I'm sorry for everyone that has to read through this :(
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1439
Next commits, and this patchset in general, will make this patchset
obsolete, since it'll only test graphene types against each other.
If at all useful, the Euler test should be moved to graphene.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1439
Use dot products to simplify calculations. Because the 'w' column of
the matrix is always summed, use 1.f in the 'w' component of the point
vector.
Because CoglMatrix is column-major and graphene_matrix_t is row-major,
it is necessary to transpose the matrix before retrieving the rows.
When we switch CoglMatrix to be row-major, this transposition will
go away.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1439
This is pretty similar to the other conversions, except we need to
store the matrix flags before operating on it, and update it using
this old value after. That's because cogl_matrix_init_from_array()
marks the matrix as entirely dirty, and we don't want that.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1439
At this point, we are still only changing CoglMatrix APIs internally, and
it should still produce the same output as before.
To achieve this, using graphens matrix implementation, we need to exploit
some knowledge about conventions used in Cogl and graphene respectively.
In Cogl, transformation matrices are equivalent to those of affine
transformation matrices. The convention used by graphene, however, is to
operate on matrices that are transposed compared to their affine
counterparts.
So for example, let's say we want to multiply the affine matrices A and B,
to get C.
A × B = C
The first step is to convert A and B to graphene matrices. We do this by
importing the floating point array, importing it directly using graphene.
Cogl exports its matrix to a column major floating point array. When we
import this in graphene, being row major, we end up with the same matrix,
only transposed.
Cogl Graphene
A <===> Aᵀ
B <===> Bᵀ
We then multiply these imported matrices in reverse
Bᵀ × Aᵀ
which in turn, due to ABᵀ = BᵀAᵀ, gives us
Bᵀ × Aᵀ = (A × B)ᵀ
Our original goal was to find C, thus we know that
A × B = C
That means we can shuffle things around a bit.
A × B = C
Bᵀ × Aᵀ = (A × B)ᵀ
Bᵀ × Aᵀ = Cᵀ
With the same conversion as done when going from Cogl to graphene, only
the other way around, we still end up effectively transposing the matrix
during the conversion.
Graphene Cogl
Cᵀ <===> C
Thus when converting Cᵀ to Cogl, we in fact end up with C.
(Explanation authored by Jonas Ådahl)
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1439
Graphene provides skewing as part of graphene_matrix_t API, and it'll
be easier for the transition to just expose similar API surfaces.
Move the matrix skew methods to CoglMatrix.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1439
Writing tests' output to a log file makes them difficult to debug when
the test might be running on an autobuilder or CI system where only
stdout/stderr are recorded. This is particularly troublesome if a
failure is only reproducible on a particular autobuilder.
Recent Automake versions have the convention that detailed output from
failing tests is written to stdout/stderr, not just to log files, when
the VERBOSE environment variable is set; borrow that convention as a
trigger for producing detailed test output.
This was originally cogl!14, but applies equally to mutter's fork of cogl.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1273
Signed-off-by: Simon McVittie <smcv@debian.org>
This resolves a couple of FIXMEs. The FIXME comments were right in
stating that not *all* journals needed flushing, only the one we
are trying to put on screen needs flushing.
However we can't eliminate all flushes because the winsys swap calls
that follow go directly into OpenGL which knows nothing about cogl
journalling. So the journal *must* be flushed before the swap, to give
OpenGL the correct state.
P.S. If this turns out to cause any bugs then the next best answer is
to just remove the FIXME comments. Because flushing is still the right
thing to do.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1362
Even when a direct client buffer has a compatible format, stride and
modifier for direct scanout, drmModePageFlip() may still fail sometimes.
From testing, it has been observed that it may seemingly randomly fail
with ENOSPC, where all subsequent attempts later on the same CRTC
failing with EBUSY.
Handle this by falling back to flipping after having composited a full
frame again.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/1410
When the CoglRenderer didn't set the DMA buffer constructor vfunc, we
return NULL. What we didn't do was set the error, meaning the caller
would crash if it tried to look up why DMA buffer construction failed.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1430
This will be used when screencasting monitors so that if
there's scanout in place, it'll still be possible to blit
it to a PipeWire-owned framebuffer, and stream it.
Add a new 'blit_to_framebuffer' vfunc to CoglScanout, and
implement it in MetaDrmBufferGbm.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1421
In certain situations it's desirable to keep pipelines around for
the whole lifetime of the session. In order to not leak them and
properly clean them up on shutdown, introduce a new mechanism to
create named pipelines that are bound to their correstponding
context and may be used across file boundries.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1372
In order to support the DRM formats DRM_FORMAT_ABGR16161616F and
friends, as well as the wl_shm formats WL_SHM_FORMAT_ABGR16161616F and
friends, cogl needs to have knowledge about said formats too.
We don't have a software implementation of the half point data types
however, so the pack/unpack methods remain unimplemented. We don't need
them for now, so it's not crucial that we add them.
For the GLES2 driver, currently only two formats are supported, and
since we don't currently have pack/unpack implementations, the other
formats will for now remain unsupported, until we have a half float
implementation.
https://gitlab.gnome.org/GNOME/mutter/merge_requests/804
In the case of indirect rendering like the first frame to use mutter's
background wallpaper:
Texture_A -> FBO_B (Texture_B) -> FBO_C (screen)
we would be trying to render the contents of both FBO_B and FBO_C in
the same flush, before the contents of Texture_A had made it to FBO_B.
So when FBO_C wants to use mipmaps of Texture_B they didn't exist yet
and appeared all black. And the blackness would remain for subsequent
frames as cogl has now decided the mipmaps of FBO_B are no longer
"dirty" and don't need refreshing:
FBO_B (Texture_B) (mipmaps_dirty==FALSE but black) -> FBO_C (screen)
We must flush FBO_B before referencing Texture_B for use in rendering
FBO_C. This only happens when Texture_A changes (e.g. when the user
changes their background wallpaper) so there's no ongoing performance
penalty from this flush.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1347
The native backend had a plain counter, and the X11 backend used the
CoglOnscreen of the screen; change it into a plain counter in
ClutterStageCogl. This also moves the global frame count setting to the
frame info constuctor.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1285
We currently have mutter set a global frame counter on the frame info in
the native backend, but in order to do this from clutter, change the
frame info construction from being implicitly done so when swapping
buffers to having the caller create the frame info and passing that to
the swap buffers call.
While this commit doesn't introduce any other changes than the API, the
intention is later to have the caller be able to pass it's own state
(e.g. the global frame count) along with the frame info.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1285
The property is deprecated and the current implementation simply
redirects it to ClutterActor::background-color, so remove it.
Also update the tests to set the background color directly.
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1332
