The GL format and type are already recorded with pixel_format_info, use
that instead of a switch on Pixman formats.
Less special-casing, less dependency on Pixman formats.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Everywhere we are standardising to drm_fourcc.h pixel format codes, and
using struct pixel_format_info as a general handle that allows us to
access the equivalent format in various APIs. In the name of
standardisation, convert weston_compositor::read_format to
pixel_format_info.
Pixman formats are defined CPU-endian, while DRM formats are defined
always little-endian. OpenGL has various definitions. Correctly mapping
between these when the CPU is big-endian is an extra chore we can
hopefully offload to pixel-formats.c.
GL-renderer read_format is still defined based on Pixman format, because
of the pecualiar way OpenGL defines a pixel format with
GL_UNSIGNED_BYTE. That matches the same Pixman format on big-endian but
not the same drm_fourcc.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
It is only enabled by a debug key binding, currently not tested at all,
and is seems it doesn't really work, so let's remove it. This also
removes it from the man page.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
By moving the application of view_alpha after pre-multiplication we can
simplify main() considerably.
The cost is that for straight-alpha input or color_pipeline() we might
be doing three multiplications more than before. However,
a) the cost of running color_pipeline() probably dominates anyway, and
b) to get straight-alpha input you have to use a future Wayland
extension that probably won't be advertised without color management.
So we keep the optimization for the simple case (no color management)
while potentially incurring a small cost on the heavy case (with color
management).
Thanks to Pierre-Yves Mordred for the inspiration in
https://gitlab.freedesktop.org/wayland/weston/-/merge_requests/889#note_1411774
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Now that we have the if-else ladder to call color_pipeline() only when
necessary, and since only color_pipeline() needs undo-premult, move
undo-premult into color_pipeline().
This is a small step towards improving code readability.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
We always talk about "view alpha", so the name variable in the fragment
shader the same. Now it's clear without the comments, making the code
easier to read overall.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This patch makes sure we have a gl_buffer_state present when using
direct-display protocol extensions (which forbids any GL imports, and
assumes a direct path with the display unit to perform a KMS import).
Without this patch we would basically have no gl_buffer_state at repaint
time because we never manged to create one, as direct-display code path
will return much early.
Partially fixes gitlab.freedesktop.org/wayland/weston/-/issues/621.
Suggested-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
When color management is disabled, the fragment shader was still first
ensuring straight alpha and then immediately just going back to
pre-multiplied. This is near-impossible for a shader compiler to
optimize out, I guess because of the if-statement to handle division by
zero. Having view alpha applied in between certainly didn't make it
easier.
That causes extra fragment computations that are unnecessary. In the
issue report this was found to cause a notable performance regression.
Fix the performance regression by introducing special-case paths for
when straight alpha is not needed. This skips the unnecessary
computations.
Fixes: https://gitlab.freedesktop.org/wayland/weston/-/issues/623
Fixes: 9a6a4e7032
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Obviously the first allocation is always leaked, there is a second
zalloc() right below. Fix the leak.
Found by code inspection.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
It's three planes, not two.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Fixes: 8b167a1703 ("gl-renderer: Store EGL buffer state in weston_buffer")
There's just no good reason to do this.
The query entrypoints already tell us if we need to use
GL_TEXTURE_EXTERNAL_OES for a particular format/modifier. We also have
RGB -> YUV fallbacks which should be able to work well with TEXTURE_2D.
TEXTURE_EXTERNAL pessimises quite hard, forcing GPU-side reloads as well
as bad filtering. Allowing multi-planar formats to use TEXTURE_2D should
thus result in performance and quality improvements.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Now that we can pull everything we need from pixel-formats, go one step
further and reuse the same YUV format descriptors we use to emulate
dmabuf/EGLImage imports for SHM.
This eliminates all special-case YUV/SHM handling.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Add a new hide_from_clients flag which, if set, specifies that the
format is only for internal information and processing, and should not
be advertised for clients.
This will be used for formats like R8 and GR88, which are not useful for
client buffers, but are used internally to implement YUV -> RGB
conversion.
Signed-off-by: Daniel Stone <daniels@collabora.com>
We support this as an explicit YUV fallback path in gl-renderer's dmabuf
EGLImage import path, so might as well support it in the SHM path, given
it's just YUV420 with no subsampling.
Signed-off-by: Daniel Stone <daniels@collabora.com>
If we're doing partial uploads from SHM buffers, we need to use the
vertical subsampling factor rather than the horizontal for secondary
planes.
Signed-off-by: Daniel Stone <daniels@collabora.com>
pixel-formats already stores the gl_format, at least for single-planar
formats; use that instead of storing our own copies.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Instead of checking for each format whether we need compatibility
workarounds for GL implementations not supporting ES3.x or when
GL_EXT_texture_rg isn't present, have each format declare the ideal case
and fix it up later.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Rather than checking all the pixel-format components which are currently
duplicated inside gl-renderer, just check for equality of the pixel
format itself, which will become useful as we remove some of the
duplicate content.
This means that the texture storage will now be reallocated when clients
switch between pixel formats which could've had compatible GL storage
(e.g. XRGB <-> ARGB) on the same surface. However this does not seem
like a case worth optimising, and simplifies the code somewhat.
Signed-off-by: Daniel Stone <daniels@collabora.com>
We've got a nice shiny ARRAY_COPY macro, so use it rather than memcpy or
hand-unrolled assignments.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Log EGL features similar to how GL ES features are logged: listing just
the ones weston tests for.
This replaces some log messages from gl-renderer.c that become
redundant or belong with EGL better.
has_native_fence_sync and has_wait_sync are not logged, because missing
them already logs warnings.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Feels like this might be nice to log.
The failure case is not fatal, so say it's a warning only.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This is a human readable replacement for printing out the list of all
available GL extensions that doesn't happen anymore by default.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Print all EGL and OpenGL extension lists into a new log scope
"gl-renderer" instead of the usual log.
These lists cluttered the log while they were very rarely actually
useful. Sometimes they might be interesting, so make them still
available through the new log scope.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Plumb struct gl_renderer all the way through to
gl_renderer_log_extensions(). In the future, the extension lists will be
printed into a debug scope specifically, and it will get the debug scope
from gr.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Now that the gl_buffer_state owns everything related to buffers, move
the textures from there rather than living on the surface, to join the
EGLImage and/or SHM params.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Now that EGLImages are strongly associated with a gl_buffer_state, which
has a lifetime strictly bounded by a weston_buffer, we don't need to
have an egl_image wrapper having its own separate refcounting anymore.
Signed-off-by: Daniel Stone <daniels@collabora.com>
... apart from SHM.
EGL and dmabuf buffers already have a gl_buffer_state created for them
when we first attach the weston_buffer. By turning
gl_surface_state::buffer into a pointer, we can just reference rather
than inline the gl_buffer_state.
SHM buffers are special, in that we don't keep individual copies of them
within the GL renderer. Instead, the GL surface has a texture allocated
with a shadow copy of the most up-to-date surface content. Handle this
by allocating and destroying gl_buffer_state every time we need to
respecify textures or somehow meaningfully change the parameters.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Similarly to EGL buffers, store the gl_buffer_state for a dmabuf buffer
inside weston_buffer, rather than on the linux_dmabuf_buffer. This
slightly simplifies our gl_buffer_state handling, and will be used later
to eliminate the egl_image refcounting.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Introduce a renderer_private hook for weston_buffer, and use this to
store a copy of the gl_buffer_state for EGL buffers (i.e. non-dmabuf, via
EGL_WL_bind_wayland_display).
As part of this, we create the EGLImage along with the weston_buffer
information, and just take a reference to it each time it is attached.
If you have bisected a failure to update surface content to this commit,
it very likely means that your EGL implementation requires images to be
recreated rather than only rebound in order to have their content
updated, which is contrary to specification.
Signed-off-by: Daniel Stone <daniels@collabora.com>
At the moment, attach_shm() will modify the gl_buffer_state in place,
then compare it and see if it differs enough to require a new one. That
rather mixes up the old and new worlds, so quite explicitly build up a
shadow gl_buffer_state with variables first before we change the one
which already exists.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Now that we can reliably access buffer dimensions from weston_buffer,
and gl-renderer isn't doing strange things with buffer widths, just use
that. The renderer interface is now unused and can be deleted.
Signed-off-by: Daniel Stone <daniels@collabora.com>
This was only used for what was presumably an attempt at an
optimisation, to force the texture's pitch in pixels to match the SHM
buffer. This is really unlikely to have ever made a difference, given
the alignments GPUs demand.
Signed-off-by: Daniel Stone <daniels@collabora.com>
It's just a shadow of weston_buffer.buffer_origin, which also has a
slightly more descriptive name.
Signed-off-by: Daniel Stone <daniels@collabora.com>
If we can reuse the textures we already have, just return early, rather
than putting all the work in a large indented body.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Pekka Paalanen
Marius Vlad
Daniel Stone