rect_union_add takes a pixman_box32_t by value, and passes it along by
value to internal helpers. render_pass_mark_box_updated which is the
only caller receives the pixman_box32_t by reference, so just plumb it
through that way.
Results in a 13% improvement in CPU time when using the Vulkan renderer
on the stacked/clip200/1024 benchmarks on my machine.
Signed-off-by: Kenny Levinsen <kl@kl.wtf>
Similar to what we have already done for gles2. To simplify things we
use the staging ring buffer for the vertex buffers by extending the
usage bits, rather than introducing a separate pool.
Signed-off-by: Kenny Levinsen <kl@kl.wtf>
We are spending quite significant CPU time walking through the clip
rects, taking a pixman box, converting it to a wlr box, intersecting it
and ultimately converting it back to a pixman box before adding it to
the rect union.
Just intersect the clip region once ahead of time, and use pixman boxes
the entire way. This also makes it easy to bail early if nothing
intersects.
Gives a small 97.95% reduction in CPU time for the Vulkan renderer in
the grid/clip200/1024 benchmark.
Signed-off-by: Kenny Levinsen <kl@kl.wtf>
Implement a ring-buffer that uses timeline points to track and release
allocated spans. We add larger ring-buffers when it fills, and remove
them when they have been unused for many collection cycles.
Signed-off-by: Kenny Levinsen <kl@kl.wtf>
Before this patch the pixman renderer would use "constant padding"
for bilinear scaling which meant that the edges would either be dark
or turn transparent. The effect was most obvious when trying to scale
a single row buffer to a height like 100. The center would have the
desired color but the edges to both sides would fade into transparency.
We now use PIXMAN_REPEAT_PAD which clamps out-of-bound pixels and
seems to match the behavior of the gles2 renderer.
Uses the EXT version of VK_PIPELINE_COMPILE_REQUIRED in `vulkan_strerror` func since it requires
Vulkan 1.3, switch to VK_EXT_global_priority instead of VK_KHR_global_priority which is only
promoted to core in Vulkan 1.3 as well.
This allows using the vulkan renderer on platforms that provide all
the necessary Vulkan extensions.
Tested on a Mali G52 platform with Mesa 26.0.0 and 25.3.5, which only
support Vulkan API 1.0.
When we're reading from a DMA-BUF texture using implicit sync, we
need to (1) wait for any writer to be done and (2) prevent any
writers from mutating the texture while we're still reading. We
were doing (1) but not (2).
Fix this by calling dmabuf_import_sync_file() with DMA_BUF_SYNC_READ
for all DMA-BUF textures we've used in the render pass.
We'll need to grab textures from there in the next commit.
Also rename it to better reflect what it does: synchronize release
fences after a render pass has been submitted.
Currently, the glslang check is run every time ninja is invoked,
even with an up-to-date build directory when GLSL shaders haven't
been modified. This is due to glslang not creating any output
file: the _check file never exists so ninja keeps trying to
generate it by running the command.
Unfortunately Meson doesn't support running commands with no
outputs [1]. Create an empty output file to fix this by setting
`capture: true`.
This doesn't work out-of-the-box, because glslang prints messages
to stdout, and provides no way to change this [2]. As a result,
shader errors are not surfaced back to the user - they end up in
the _check file. Workaround this with a thin wrapper which
redirects stdout to stderr when invoking glslang.
[1]: https://github.com/mesonbuild/meson/issues/11506
[2]: https://github.com/KhronosGroup/glslang/pull/4138
Fixes use-after-free on exit of labwc running nested:
==50906== Invalid write of size 8
==50906== at 0x4A85403: wl_list_remove (wayland-util.c:57)
==50906== by 0x40BBAF9: destroy_wl_buffer (output.c:146)
==50906== by 0x40B9B4F: backend_destroy (backend.c:488)
==50906== by 0x409E96F: wlr_backend_destroy (backend.c:68)
==50906== by 0x40B78A6: multi_backend_destroy (backend.c:62)
==50906== by 0x409E96F: wlr_backend_destroy (backend.c:68)
==50906== by 0x4043DA0: server_finish (server.c:788)
==50906== by 0x403AA85: main (main.c:277)
==50906== Address 0xb4435e8 is 40 bytes inside a block of size 136 free'd
==50906== at 0x4A3E8EF: free (vg_replace_malloc.c:989)
==50906== by 0x409C954: buffer_destroy (shm.c:28)
==50906== by 0x40E96F4: buffer_consider_destroy (buffer.c:42)
==50906== by 0x40E9754: wlr_buffer_drop (buffer.c:52)
==50906== by 0x41498DA: slot_reset (swapchain.c:44)
==50906== by 0x4149933: wlr_swapchain_destroy (swapchain.c:53)
==50906== by 0x40CB1FA: wlr_output_finish (output.c:410)
==50906== by 0x40BE00B: output_destroy (output.c:957)
==50906== by 0x40CB2FC: wlr_output_destroy (output.c:436)
==50906== by 0x40B9AFC: backend_destroy (backend.c:481)
==50906== by 0x409E96F: wlr_backend_destroy (backend.c:68)
==50906== by 0x40B78A6: multi_backend_destroy (backend.c:62)
==50906== Block was alloc'd at
==50906== at 0x4A42C13: calloc (vg_replace_malloc.c:1675)
==50906== by 0x409CA84: allocator_create_buffer (shm.c:68)
==50906== by 0x409C7BA: wlr_allocator_create_buffer (allocator.c:186)
==50906== by 0x4149B80: wlr_swapchain_acquire (swapchain.c:102)
==50906== by 0x40C90DA: render_cursor_buffer (cursor.c:246)
==50906== by 0x40C93DC: output_cursor_attempt_hardware (cursor.c:303)
==50906== by 0x40C9A61: output_cursor_set_texture (cursor.c:420)
==50906== by 0x40C9738: wlr_output_cursor_set_buffer (cursor.c:352)
==50906== by 0x40F13A0: output_cursor_set_xcursor_image (wlr_cursor.c:507)
==50906== by 0x40F1B28: cursor_output_cursor_update (wlr_cursor.c:630)
==50906== by 0x40F1C2A: cursor_update_outputs (wlr_cursor.c:657)
==50906== by 0x40F1CF9: wlr_cursor_set_xcursor (wlr_cursor.c:674)
Fixes: 7963ba6a0d
("buffer: introduce wlr_buffer_finish()")
BT.1886 is different from other EOTFs in that the spec says that offset
b must be in electrical domain, so use the assumed display environment
from the specification as Lmin and Lmax then normalize bt.1886 to
produce optical luminance L in [0, 1]. See discussion at
https://gitlab.freedesktop.org/pq/color-and-hdr/-/merge_requests/63#note_3090016
Fixes d1c88e9 "render/vulkan: add linear single-subpass"
`find_or_create_render_setup` still assumed a single-buffer setup
always implied use of an srgb format
Not all eotf or eotf inverse are well defined for values outside the
intended domain, so just ignore it and clamp it away.
An alternative solution would be to use sign preserving pow here (i.e.
sign(x) * pow(abs(x), p)), however I'm not sure that makes sense or is
defined anywhere. Negative values here are likely a result of colors
being outside the gamut range, so clipping them to 0 is more correct
than mirroring from 0.
