Commit Graph

378 Commits

Author SHA1 Message Date
Quantum
15066c7345 [client] egl: handle eglQuerySurface(EGL_BUFFER_AGE_EXT) error 2021-12-14 14:35:30 +11:00
Quantum
58f83da7bb [client] egl: increase the damage box for the cursor 2021-12-02 22:07:37 +11:00
Quantum
3e9a21d3b9 [client] egl: use linear filter when not scaling
This appears to eliminate artifacts related to non-full screen paints
due to precision errors with the nearest filter.
2021-12-02 22:07:24 +11:00
Geoffrey McRae
8167ef2c4a [client] egl: make use of glsl's textureSize function 2021-11-12 07:41:59 +11:00
Geoffrey McRae
67ea8e06ba [client] egl: use alloca instead of typecasted char array 2021-10-24 22:57:21 +11:00
Quantum
ad6e3f96e6 [client] egl: reset damage on renderer restart 2021-10-23 18:38:21 +11:00
Geoffrey McRae
fe712b7ec9 [client] egl: damage the full screen if the frame format changes 2021-10-22 18:51:49 +11:00
Geoffrey McRae
dca5da02a0 [client] egl: fix undefined behaviour with zero size array 2021-10-20 13:34:16 +11:00
Quantum
2858ad3f7e [client] egl: fallback when EGL_RENDER_BUFFER fails
This allows the client to work when the OpenGL implementation fails if
EGL_RENDER_BUFFER is passed, printing a warning. This should fix issues
with Nvidia proprietary drivers on Wayland.
2021-10-20 11:31:00 +11:00
Quantum
9ddd260b22 [client] egl: log glCheckFramebufferStatus error 2021-10-20 11:30:40 +11:00
Geoffrey McRae
bc34dc9e24 [client] egl: blame NVIDIA if dmabuf fails on NVIDIA hardware
NVIDIA still do not implement a complete/working DMABUF implementation
yet advertise support. Best to tell the public to complain to NVIDIA
instead of assuming it's a LG bug or an issue with their system.
2021-10-20 11:28:29 +11:00
Geoffrey McRae
edc9825c04 [client] egl: check for invalid desktop render dimensions 2021-10-20 11:17:29 +11:00
Geoffrey McRae
70a751b58d [client] egl/filters: bypass setup/prepare logic if disabled/inactive
filters are now not run if `egl_filterSetup` returns false, as such we
do not need additional `enable` checks in `prepare` or `run` and we can
bypass the filters even earlier if they are not enabled reducing load.
2021-10-20 10:57:53 +11:00
Quantum
fc037ccc95 [client] egl: handle filter setup returning false
As discussed, this should just skip the filter as if it's disabled.
2021-10-20 10:48:18 +11:00
Quantum
74418106de [client] egl: skip downscale filter setup when not enabled
Since this->prepared will never be set to true unless the filter is
enabled, this results in the framebuffer setup being done every frame
for no reason, causing a lot of texture reallocations.
2021-10-20 10:47:59 +11:00
Quantum
8dcf7af791 [client] egl: fix parentheses in IDX_AGO definition
The old definition could break if complex expressions were passed in
as arguments.
2021-10-17 06:10:30 +11:00
Quantum
2d858da0f1 [client] egl: fix egl_bufferAge naming 2021-10-17 06:10:30 +11:00
Quantum
ee211803e4 [client] egl: fix letterbox rendering
Add additional half pixel in all directions to ensure erasure of
content in the letterbox area under all circumstances.
2021-10-17 06:10:30 +11:00
Quantum
148ab0278e [client] egl: add debug options to disable damage tracking
This should aid in figuring out the source of damage tracking bugs.
2021-10-14 17:01:48 +11:00
Quantum
d60dcb718b [client] cmake: correctly detect non-gawk awks
Forgetting NAMES meant that cmake was searching for gawk only.
2021-10-14 17:01:37 +11:00
Quantum
4a76401c34 [client] egl: make the last eglDestroyImage call indirect
Missed this one in the last PR.
2021-09-29 18:28:07 +10:00
Quantum
072c54977e [client] egl: use eglCreateImage and eglDestroyImage indirectly
The dmabuf path is optional, so we shouldn't require those functions to
link our program.
2021-09-29 17:48:50 +10:00
Quantum
1b58f2592c [client] egl: make filters damage aware
This saves a lot of GPU power for partial updates. Running testufo with
lanczos downscaling and FSR upscaling consumed over 90 W, but with this
commit, consumed only 75 W.
2021-09-04 13:30:24 +10:00
Quantum
e9bf225c75 [client] egl: remove useless RenderStep struct 2021-09-04 13:28:49 +10:00
Quantum
ac926e4458 [client] egl: support debug contexts on older EGL versions 2021-09-04 13:24:45 +10:00
Quantum
26d2feb4d8 [client] egl: remove warning disabling in ffx.c 2021-09-04 13:24:34 +10:00
Quantum
977a4f6323 [client] egl: remove useless comment in filter_downscale.c 2021-08-31 20:14:39 +10:00
Quantum
f4df690d9f [client] egl: fix preset tooltip always showing 2021-08-30 18:47:48 +10:00
Quantum
239cb6a92b [client] egl: add separator after presets 2021-08-30 18:45:21 +10:00
Quantum
876a4125bf [client] egl: correctly select the preset when saving over it 2021-08-30 18:45:21 +10:00
Quantum
ffbf6cd8b4 [client] egl: make preset UI more intuitive
This commit makes selecting a preset load it, and changes "Create preset" to
"Save preset as...".
2021-08-30 18:45:21 +10:00
Quantum
39e42ba735 [common] option: change option_dump to option_dump_preset
This new function dumps all options marked as preset instead of dumping
individual sections. This should allow filter options to not be all grouped
into the [eglFilter] section.
2021-08-30 18:32:16 +10:00
Quantum
3345ff8448 [client] egl: mark all preset-only options as such 2021-08-30 18:21:54 +10:00
Quantum
8a2ae6860e [client] egl: warn when attempting to save without preset selected 2021-08-30 18:21:54 +10:00
Quantum
1717555187 [client] egl: implement preset deletion 2021-08-30 18:21:54 +10:00
Quantum
dc27638025 [client] egl: implement loading and saving of filter order 2021-08-30 18:21:54 +10:00
Quantum
c85cc7d668 [client] egl: show preset errors as modal dialogs 2021-08-30 18:21:54 +10:00
Quantum
311f7241c6 [client] egl: implement preset loading/saving logic 2021-08-30 18:21:54 +10:00
Quantum
b7b93f624c [client] egl: implement options loading/saving for downscale 2021-08-30 18:21:54 +10:00
Quantum
084837b936 [client] egl: add loadState and saveState for filters 2021-08-30 18:21:54 +10:00
Quantum
4adb425337 [client] egl: add UI for presets list 2021-08-30 18:21:54 +10:00
Quantum
5225d2e97f [client] egl: fix framebuffer leaking textures 2021-08-28 19:17:24 +10:00
Quantum
e6df0acad9 [common] vector: eliminate double allocation when possible
This commit creates two constructor/destructor pairs for vector:
* vector_alloc/vector_free dynamically allocates the vector itself
* vector_create/vector_destroy uses existing Vector objects
2021-08-28 19:17:15 +10:00
Quantum
ceff9dca9b [client] egl: simplify filter moving logic with memmove
This avoids duplicating the entire array of filters.
2021-08-25 05:35:50 +10:00
Quantum
7c7eff8dba [client] egl: make texture_dmabuf.c use vector
This replaces the custom memory management code.
2021-08-24 22:10:36 +10:00
Geoffrey McRae
b118c3b681 [client] egl: implement nicer drag & drop re-ordering of filters 2021-08-24 22:05:46 +10:00
Quantum
e5e76d784e [client] egl: allow postprocessing filters to be reordered by dragging 2021-08-22 21:36:13 +10:00
Quantum
99761b195f [client] egl: switch postprocessing filters to use vectors
This will allow them to be reordered much more easily.
2021-08-22 21:36:13 +10:00
Quantum
9aa0d3ddab [client] egl: fix context creation on EGL 1.4
EGL_CONTEXT_OPENGL_DEBUG is only defined in EGL 1.5, and therefore, we should
not be passing it on older versions of EGL.
2021-08-19 21:28:56 +10:00
Quantum
429620c48b [client] egl: dynamically import glBufferStorageEXT
On some implementations (e.g. llvmpipe), the function can only be queried via
eglGetProcAddress.
2021-08-19 21:28:45 +10:00
Quantum
bb91b41c64 [client] egl: look at 3x3 around the pixel instead of 4x4
Using 4x4 means that some pixels will be outside of the lanczos window. The
ideal lanczos function should in fact be zero in those areas, so we shouldn't
waste time processing those pixels.

I can't notice any difference in the results.
2021-08-19 15:52:44 +10:00
Quantum
520460669c [client] egl: set gl_Position.z in cursor vertex shader 2021-08-19 12:24:55 +10:00
Quantum
81c38e825c [client] remove all casts around malloc
The cast is unnecessary in C and should be removed to avoid clutter.
2021-08-16 16:26:58 +10:00
Quantum
fd4a4114e6 [client] egl: pad areas of the desktop repainted to cover overlays
We pad the screen coordinates and then convert to desktop coordinates,
so that the padding will always be a pixel wide on screen.
2021-08-16 16:26:18 +10:00
Quantum
104141eec1 [client] use correct argument order for calloc 2021-08-16 16:25:59 +10:00
Quantum
b7d3bbbd82 [client] egl: use standard-compliant way of EGL detection
According to the documentation for eglQueryString:

> EGL_BAD_DISPLAY is generated if display is not an EGL display connection,
> unless display is EGL_NO_DISPLAY and name is EGL_EXTENSIONS.

Therefore, we should check EGL by doing:

    eglQueryString(EGL_NO_DISPLAY, EGL_EXTENSIONS)

Indeed, the old way of eglQueryString(EGL_NO_DISPLAY, EGL_VERSION) works on
libglvnd but not using mesa's libEGL.so directly.

Also added a warning to make it more obvious that EGL is not available.
2021-08-16 16:25:48 +10:00
Tudor Brindus
14ad83c6b8 [client] use variable-relative sizeof where possible 2021-08-16 16:22:55 +10:00
Quantum
8f5afe1848 [client] egl: clamp sharpness settings in filters
While the slider does not allow you to get out of range by dragging,
the user could still type in out of range values, so we clamp the values.
2021-08-15 18:01:03 +10:00
Quantum
36073586e7 [client] egl: add tooltip about Ctrl+Click on sharpness sliders
With the new keymap feature, we are now able to properly support letting
the user enter exact values into the sliders. This commit adds a tooltip
to help the user discover this feature.

Note that this currently only works on Wayland. The X11 backend will need
to call app_handleKeyboardModifiers.
2021-08-15 18:01:03 +10:00
Quantum
bbd173000f [client] egl: clamp downscale factor range 2021-08-14 14:57:33 +10:00
Quantum
96738ab9d0 [client] egl: make downscale filter use text input widget
The major/minor pixel size hack is too confusing. This commit replaces
that with a text input and a slider.
2021-08-14 14:44:26 +10:00
Quantum
5f3bd778c0 [client] egl: add debug prints for renderStartup errors 2021-08-14 12:20:12 +10:00
Quantum
f66486b0c7 [client] egl/downscale: implement filter switching 2021-08-14 12:19:50 +10:00
Quantum
2c02e6c4a0 [client] egl: add linear downscale filter shader 2021-08-14 12:19:50 +10:00
Quantum
94de061587 [client] egl: implement lanczos filter shader 2021-08-14 12:19:50 +10:00
Quantum
16adbab5d4 [client] all: remove needless initalization 2021-08-14 12:19:07 +10:00
Quantum
579f998519 [client] all: replace assert with DEBUG_ASSERT 2021-08-14 12:19:07 +10:00
Quantum
85a96d1e06 [client] all: use DEBUG_UNREACHABLE instead of assert
Due to the way assert is defined in standard C, compilers in release mode
will not treat it as unreachable. This explains a lot about those pesky
uninitialized variable bugs, actually.
2021-08-14 12:19:07 +10:00
Quantum
8c18817e2d [client] egl: don't generate mipmaps in downscale filter
When using DMABUF, the mipmaps can cause driver hangs and crashes.
2021-08-12 17:03:18 +10:00
Geoffrey McRae
117e88c240 [client] egl: add new downscale filter 2021-08-12 15:54:16 +10:00
Geoffrey McRae
6387bf2d2e [client] rework the configuration overlay to allow for tabs 2021-08-12 09:04:45 +10:00
Geoffrey McRae
fe6339fc77 [client] egl: re-order CAS to before FSR for better results 2021-08-12 06:56:16 +10:00
Quantum
3f8c7c8d0d [client] egl: fix buffer overflow in desktop_rects module
The module has been changed to support variable amount of rectangles,
so we should just allocate a VLA.
2021-08-11 21:23:13 +10:00
Quantum
543c97987b [client] egl: remove needless precision quantifiers
We simply use precision mediump float; for everything. We don't actually
need highp anyways, and we don't use it for stuff like CAS or FSR.
2021-08-11 20:47:46 +10:00
Geoffrey McRae
06da52acfc [client] egl/fsr: release consts when no longer needed 2021-08-11 20:47:03 +10:00
Geoffrey McRae
5a2f34d71c [client] egl/cas: release consts when no longer needed 2021-08-11 20:42:56 +10:00
Quantum
8b2db071d8 [client] egl: precompute CAS filter constants on CPU 2021-08-11 20:38:42 +10:00
Quantum
3a1a9121eb [client] egl: make FSR filter show inactive in config when disabled 2021-08-11 20:17:17 +10:00
Quantum
f80b67bc50 [client] egl: precompute FSR filter constants on CPU 2021-08-11 20:16:39 +10:00
Quantum
fe823b6172 [client] egl: display FSR equivalent quality mode
This also displays a tooltip to explain that quality can be changed by
altering guest resolution and also show the resolutions needed to achieve
each quality mode.
2021-08-11 20:05:27 +10:00
Quantum
c4c60fd330 [client] egl: update FSR filter state upon resolution change 2021-08-11 20:05:27 +10:00
Geoffrey McRae
5a5b867c73 Revert "[client] egl: make FSR detect an input size change and activate if valid"
This reverts commit 73f125dcc7.
2021-08-11 20:05:07 +10:00
Geoffrey McRae
73f125dcc7 [client] egl: make FSR detect an input size change and activate if valid 2021-08-11 20:02:29 +10:00
Quantum
7e982a6658 [client] util: replace util_mergeOverlappingRects with common version 2021-08-11 19:01:52 +10:00
Geoffrey McRae
f0ea882165 [client] egl: cleanup texture filtering/post-processing 2021-08-11 18:53:36 +10:00
Geoffrey McRae
f78154d282 [client] egl: fixes to shader post-process pipeline 2021-08-11 06:31:01 +10:00
Quantum
cd5ecf3e5a [client] egl: don't erase damage when invalidating whole window 2021-08-11 02:43:08 +10:00
Quantum
a850a1b51b [client] egl: implement C wrappers for FidelityFX constant computation 2021-08-11 02:42:55 +10:00
Quantum
c1a362f8d3 [client] egl: handle \r character when processing #includes 2021-08-11 02:41:54 +10:00
Geoffrey McRae
127d3acd96 [client] egl: use a texel based version of textureGather for FSR 2021-08-10 14:21:46 +10:00
Quantum
ccee347740 [client] egl: don't define FSR_RCAS_F in ffx_fsr1_rcas.frag 2021-08-10 13:57:23 +10:00
Geoffrey McRae
c3a143732c [client] egl: cosmetics 2021-08-10 13:46:48 +10:00
Geoffrey McRae
dc0b3a8d45 [client] egl: rework post process filters and add AMD FXR 2021-08-10 13:46:48 +10:00
Quantum
3b751a2017 [client] egl: perform full copy for framebuffer textures after resize
This prevents the code from using damage rectangles that are no longer on the
screen, causing an out-of-bounds write.
2021-08-10 13:42:25 +10:00
Geoffrey McRae
230ce81eb8 [client] egl: allocate space for the initial texture dimensions
This fixes a buffer overrun when writing to the dimensions array
2021-08-10 09:41:56 +10:00
Geoffrey McRae
e707f9d933 [client] egl: enable ffxCAS if disabled and the sharpness is changed 2021-08-10 07:56:24 +10:00
Geoffrey McRae
64ed383128 [client] egl: re-process the texture and invalidate if a setting changed 2021-08-10 07:51:23 +10:00
Quantum
685499a0e0 [client] egl: prefer gawk and mawk when building shaders
We'd rather use known versions of awk if possible for ease of troubleshooting.
2021-08-10 06:17:38 +10:00
Quantum
705250f23d [client] egl: correct assign to gl_Position in basic.vert
gl_Position is expected to be using homogeneous coordinates, which requires
w to be a coordinate scale factor, usually 1.0. z should also be set in order
for depth to be well-defined. Therefore, we should set gl_Position.zw to
vec2(0.0, 1.0).
2021-08-10 06:10:42 +10:00
Quantum
eb680086ef [client] egl: correctly use flexible array members for BindData
Array size of 0 is a gcc extension, the standard C is not declaring a size.
2021-08-10 06:10:42 +10:00