Commit Graph

578 Commits

Author SHA1 Message Date
Timm Bäder
ebbb729aea gl renderer: Linear gradient state tracking 2020-07-28 05:34:12 +02:00
Timm Bäder
dded144313 gl renderer: Assert that we have program state
This is never a problem in reality, just for static analysis.
2020-07-28 05:34:12 +02:00
Timm Bäder
3f59566729 gl renderer: Repat node child bounds can't be NULL 2020-07-28 05:34:12 +02:00
Timm Bäder
e0420bae09 gl renderer: Remove dead code
We aren't using n_corners anyway
2020-07-28 05:34:12 +02:00
Timm Bäder
9cdf9847c2 gl: Add a gdk_gl_context_has_debug accessor
Check if the label_object_ etc. functions are gonna do anything at all.
The g_type_name_from_instance calls keep poppoing up in profiles.
2020-07-28 05:34:12 +02:00
Timm Bäder
0ce32cd4b5 gl renderer: Save rectilinearity in clip stack
So we don't have to repeatedly call it in pop_clip()
2020-07-28 05:34:12 +02:00
Timm Bäder
5a9b54df96 gl renderer: More color matrix state tracking 2020-07-28 05:34:12 +02:00
Timm Bäder
d7ba281c15 gl renderer: Add contains_rect helper 2020-07-28 05:34:12 +02:00
Timm Bäder
c6c87b97c2 gl renderer: Add own graphene_rect_intersects 2020-07-28 05:34:11 +02:00
Timm Bäder
2fa9dddea6 gsk: Remember border node uniformity
Lots of border nodes have all the same width and the same color.
Renderers might be able to simplifty that, so keep an extra bit per
border node.
2020-07-28 05:34:11 +02:00
Timm Bäder
72d4a83af8 gl renderer: Moar state tracking
Track what we really need to send for inset shadows, which are used
as a border replacement in many cases.

Fishbowl says I can draw around 200-300 more switches per frame like
this too.
2020-07-28 05:34:11 +02:00
Timm Bäder
fcf502a686 gl renderer: Fix scrolledwindow undershoot rendering 2020-07-28 05:34:11 +02:00
Timm Bäder
37d7ab88dc gl renderer: Avoid 9-slicing things that can't be
The outline needs to be large enough so changing the size of the outline
does not affect the resulting blurred texture.

Fixes #2776
2020-07-17 05:49:11 +02:00
Timm Bäder
6cde64072d gl renderer: Remove outdated comment 2020-07-17 05:49:11 +02:00
Timm Bäder
6a60b0f591 gl renderer: Try to intersect rounded clips more often 2020-07-17 05:49:11 +02:00
Timm Bäder
339d9f05e4 gl renderer: Remove unused flag 2020-07-17 05:49:11 +02:00
Timm Bäder
34008e6aab gl renderer: Ignore rounded clip that only increase the clip
This fixes the widget factory rendering too much.

In the widget-factory, we generally have a pretty small update area (two
spinners and a progressbar). We take the extents of that as a update
area and inital clip.

However, the first clip node we see is from the toplevel window, which
essentially increases the clip again to almost the entire window.

Fix that by ignoring such cases.
2020-07-17 05:49:11 +02:00
Timm Bäder
21f9148155 gl renderer: Fix cross-fade nodes with invisible end child
Only loading the vertex data doesn't draw anything.
2020-06-27 10:51:06 +02:00
Timm Bäder
d4f0593b37 gl renderer: Handle 0/1 progress in crossfade nodes 2020-06-27 10:51:06 +02:00
Timm Bäder
d5bf3c2cd1 gl renderer: Improve clipping code
don't render a clip to a texture if the new clip does not intersect with
any of the corners of the currently rounded clip.

Fixes #2770
2020-05-24 15:59:05 +02:00
Timm Bäder
094788f1a3 gl renderer: Refactor render_clipped_child 2020-05-24 15:59:05 +02:00
Timm Bäder
23e68c0988 gl renderer: Cache whether clips is rectilinear 2020-05-24 15:59:05 +02:00
Timm Bäder
930d779eb3 gl renderer: don't pretend all transform nodes support being transformed
It doesn't work in the general case. The transform handling could use a
general overhaul though.

Fixes #2773
2020-05-22 21:29:10 +02:00
Timm Bäder
f3ca814f3d gl renderer: Fix some off-by-one and rounding errors in the shadow code
Take 24

Fixes #2759
2020-05-22 21:29:10 +02:00
Timm Bäder
1aa86d23f4 gl renderer: Fix blurred outset shadow slicing
Take 23.

Fixes #2759
2020-05-21 15:32:12 +02:00
Timm Bäder
01c0acb707 gl renderer: Apply scaled clip when rendering offscreen clipped child 2020-05-17 15:30:28 +02:00
Timm Bäder
007de568c9 gl renderer: Fix an uninitialized value 2020-05-17 08:07:11 +02:00
Timm Bäder
b9ed957af3 gl renderer: Fix nested rounded clip rendering
If the inner clip intersects with the corners of the outer clip, we
potentially need a texture. We should add more fine-grained checks for
this in the future though.

Test case included.
2020-05-13 08:07:55 +02:00
Timm Bäder
b98ad549c2 gl renderer: Only destroy shader programs with ids > 0
We may return -1 as an error return value.
2020-05-10 09:30:15 +02:00
Matthias Clasen
c920796967 gsk: Fall back to cairo if compiling shaders fails
There was already some fallback in place here, but
in other places we were just asserting.
2020-04-17 13:42:33 -04:00
Emmanuele Bassi
d701a89281 Turn GskRenderNode into a derivable type
Language bindings—especially ones based on introspection—cannot deal
with custom type hiearchies. Luckily for us, GType has a derivable type
with low overhead: GTypeInstance.

By turning GskRenderNode into a GTypeInstance, and creating derived
types for each class of node, we can provide an introspectable API to
our non-C API consumers, with no functional change to the C API itself.
2020-04-08 15:40:15 +01:00
Benjamin Otte
e356d59a92 build: Add -Wnull-dereference
Sprinkle various g_assert() around the code where gcc cannot figure out
on its own that a variable is not NULL and too much refactoring would be
needed to make it do that.

Also fix usage of g_assert_nonnull(x) to use g_assert(x) because the
first is not marked as G_GNUC_NORETURN because of course GTester
supports not aborting on aborts.
2020-03-05 08:14:37 +01:00
David Hogan
0969d06925 Fix detection of OpenGL 3.3 core GL_ARB_timer_query.
Some systems (notably macOS) will not allow enumeration of an extension that has been promoted to core OpenGL for context in use. This change assumes that GL_ARB_timer_query is available on OpenGL 3.3+.

I could not find definitive information on whether GL_ARB_debug_output or GL_KHR_debug have been added to core. Other extensions in use were addressed by https://gitlab.gnome.org/GNOME/gtk/merge_requests/1422 .
2020-02-14 05:41:01 +11:00
Matthias Clasen
a66ed7c461 gsk: Fix the compiler warning differently 2020-02-13 10:42:21 -05:00
Matthias Clasen
fb84fe7065 Fix flickery hover
Commit 47c44644b1 was a bit overzealous in fixing
compiler warnings. We still need to call collect_textures,
even if we don't need the number that it returns.
2020-02-12 22:45:24 -05:00
Alexander Larsson
2890cd849f profiler: Add _end_mark() version of _add_mark()
These don't take a duration, instead they call g_get_monotonic_time() to
and subtract the start time for it.

Almost all our calls are like this, and this makes the callsites clearer
and avoids inlining the clock call into the call site.
2020-02-12 11:25:34 +01:00
Alexander Larsson
01d5ad2056 profiler: Make profiler-is-running a macro
When we use if (GDK_PROFILER_IS_RUNNING) this means we get an
inlined if (FALSE) when the compiler support is not compiled in, which
gets rid of all the related code completely.

We also expand to  G_UNLIKELY(gdk_profiler_is_running ()) in the supported
case which might cause somewhat better code generation.
2020-02-12 11:05:01 +01:00
Alexander Larsson
cc643df88b Convert all profiler times from nsec to usec
usec is the scale of the monotonic timer which is where we get almost
all the times from. The only actual source of nsec is the opengl
GPU time (but who knows what the actual resulution of that is).

Changing this to usec allows us to get rid of " * 1000" in a *lot* of
places all over the codebase, which are ugly and confusing.
2020-02-12 10:44:17 +01:00
Emmanuele Bassi
042a5736fa Merge branch 'share-gl-programs' into 'master'
GskGLRenderer: Share programs between different renderers in same display

See merge request GNOME/gtk!1413
2020-02-11 16:30:16 +00:00
Matthias Clasen
720e48ec51 Merge branch 'handle-max-texture-size' into 'master'
gl: Don't require too large texture for offsreen ops

See merge request GNOME/gtk!1411
2020-02-11 16:23:02 +00:00
Alexander Larsson
7b0b3d9547 GskGLRenderer: Share programs between different renderers in same display
This is similar to how we share texture atlases. Some added complexity
in that the program state also needed to be shared, so it had to move to
the shared Programs object.

With this change realization of additional GskRenderers when opening
popups went from ~60msec to ~35 msec on average.
2020-02-11 16:58:45 +01:00
Emmanuele Bassi
47c44644b1 Silence compiler warnings in non-debug builds 2020-02-11 14:47:22 +00:00
Alexander Larsson
8fdb710c0f gl: Don't require too large texture for offsreen ops
When rendering ops to an offscreen texture we take max-texture-size
in consideration and modify the scale we use such that the required
texture does not exceed the limit.

This means some rendering will be blocky/fuzzy, but that is better
than it being clipped.
2020-02-11 13:50:35 +01:00
Timm Bäder
77e0d360ed Add pure and const annotations to various functions 2020-01-26 18:21:07 +01:00
Timm Bäder
f62535affc gl renderer: Avoid some work when rendering border nodes 2020-01-26 18:21:07 +01:00
Matthias Clasen
a297129685 Add a profiler mark around renderer realization
This takes time that is worth calling out.
2020-01-25 01:28:24 -05:00
Timm Bäder
f80a341b4f gl renderer: handle cross-fade nodes with invisible children 2020-01-24 06:19:16 +01:00
Timm Bäder
77f8245c0c gl renderer: Handle blur nodes with invisible children 2020-01-24 06:19:16 +01:00
Timm Bäder
8afdbd6ee3 gl renderer: Make render_rounded_clip_node clearer 2020-01-24 06:19:16 +01:00
Timm Bäder
ab0869ecbc gl renderer: Add builder offset correctly for non-affine modelviews 2020-01-24 06:19:16 +01:00
Timm Bäder
685288216f gl renderer: Move rect transformation to the vertex shader
No need to do this for every fragment.
2020-01-24 06:08:39 +01:00
Timm Bäder
c8fa1f19db gl renderer: Transform rounded rect on the GPU
Change the RoundedRect struct we use in our shaders so we can transform
it using (affine) matrices.
2020-01-24 06:08:39 +01:00
Matthias Clasen
1565f597c8 gl: Disambiguate profiler marks
"render" was used for both GL and widgets. Oops.
2020-01-22 16:37:48 -05:00
Timm Bäder
433c8307b2 gl renderer: Pull a local variable in the closest scope 2020-01-18 08:49:52 +01:00
Timm Bäder
dc8dedce07 gl renderer: Draw outset shadows white
and only apply the actual shadow color when we draw them from the
texture. This way we can reuse the cached shadows during color
transitions.
2020-01-18 08:49:52 +01:00
Timm Bäder
3091679ffa gl renderer: Render simple border nodes in a simple way
Roughly 80% of the border nodes are just one color and have the same
width on all sides, so we can draw them by uploading just one rect and
not four.
2020-01-18 08:49:52 +01:00
Timm Bäder
afa991752c gl renderer: Look at shadow color in the outset shadow cache
It would probably be better to not do this and always render the outline
in plain white, then later recolor it but do this for no, just for
correctness.
2020-01-18 08:49:51 +01:00
Timm Bäder
ea810f176b gl renderer: Handle repeat nodes with invisible children
We can handle this pretty easily by normally drawing the other
(visible) child node, if any.
2020-01-07 17:27:19 +01:00
Timm Bäder
6ba6f361be gltextureatlas: Packing must succeed here 2020-01-07 17:27:18 +01:00
Timm Bäder
76795ffba9 gl renderer: Sprinkle some const around 2020-01-07 17:27:18 +01:00
Timm Bäder
eebe67cf0a gl renderer: Add util to load vertex data with texture region 2020-01-07 17:27:18 +01:00
Timm Bäder
b9726901a5 gl renderer: Fix inverted dy in inset shadows 2020-01-07 17:27:17 +01:00
Timm Bäder
8b1af398a0 gl renderer: Avoid another manual rounded rect transform
We have transform_rect() for this now.
2020-01-07 17:27:17 +01:00
Timm Bäder
b8d1c3ab55 gl renderer: Don't force repeat node child offscreen 2020-01-07 17:27:17 +01:00
Timm Bäder
054d69aaf9 gl renderer: Fix a misleading comment 2020-01-07 17:27:17 +01:00
Timm Bäder
1dea6d4fc7 gl renderer: Speed up bounds transformation
We started saving the scale of the current modelview directly in the
RenderOpBuilder, so we don't need to poke the modelview stack anymore.
2020-01-07 17:27:17 +01:00
Timm Bäder
022d19a4c6 gl renderer: Add debug code for color nodes 2020-01-07 17:27:17 +01:00
Timm Bäder
ef7276c398 gl renderer: Trivial variable reordering 2020-01-07 17:27:17 +01:00
Timm Bäder
0a8ca49b5e gl renderer: transform border outline like everything else 2020-01-07 17:27:17 +01:00
Timm Bäder
c0d4a6fc81 gl renderer: Avoid copying a matrix 2020-01-07 17:27:17 +01:00
Timm Bäder
587bc82c37 gl renderer: Print debug node message 2020-01-07 17:27:17 +01:00
Timm Bäder
750dc8dbe2 gl driver: Inline function into only caller 2020-01-07 17:27:17 +01:00
Timm Bäder
db91b6dc61 gl renderer: Load vertex data directly into vertices GArray 2020-01-07 17:27:16 +01:00
Timm Bäder
44ac2d5abb gl renderer: Pass linear gradient stops to shaders directly
No need to copy them into the render ops like this.
2020-01-07 17:27:16 +01:00
Timm Bäder
f79c807645 gl renderer: Don't invert offset value twice 2020-01-07 17:27:16 +01:00
Timm Bäder
73f2167fe4 gl renderer: Avoid an invalid read
We need to copy the color here, since the program state can live across
frame boundaries.
2020-01-07 17:27:16 +01:00
Timm Bäder
f07397f4dd gl renderer: Upload GL_RGBA texture data in GLES
Colored icons are still broken this way, but at least they show up and
text works.
2020-01-07 17:27:16 +01:00
Timm Bäder
f1751f514c gltextureatlas: Use more correct debug name for texture
These aren't all glyph atlases anymore.
2020-01-07 17:27:16 +01:00
Timm Bäder
8dd7f5aefe gl renderer: Mark beginning of frame if DEBUG_OPS is set 2020-01-07 17:27:16 +01:00
Timm Bäder
398f49ad31 gl renderer: Trivial declaration reorder 2020-01-07 17:27:16 +01:00
Timm Bäder
3d260a950e gl renderer: Remove a few outdated comments 2020-01-07 17:27:16 +01:00
Timm Bäder
f31667f437 gl renderer: Remove unused translation handling 2020-01-07 17:27:15 +01:00
Timm Bäder
219493c818 gl renderer: Don't copy matrix/offset of color matrix nodes into ops 2020-01-07 17:27:15 +01:00
Timm Bäder
06f63764fb gl renderer: Initialize alpha uniform to 1.0
This makes gl-legacy work and gl-gles work except for text and icons.
2020-01-07 17:27:15 +01:00
Timm Bäder
1243174e53 gl renderer: Make RoundedRect work in gles
Which can't return struct types containing arrays. So let's revert to
the previous version but still send the rect along as a vec4[3];
2020-01-07 17:27:15 +01:00
Timm Bäder
cc909b160f gl renderer: Rewrite shader builder
Use a unified approach to write both vertex and fragment shader in the
same file.
2020-01-07 17:27:15 +01:00
Timm Bäder
d12dde07c3 gl renderer: Fix program uniform updating code
This breaks the initial uniform value, but we will fix that in a later
commit in a different way.
2020-01-07 17:27:15 +01:00
Timm Bäder
5191b6fccd gl renderer: Only send clip corners if we need to
We change the clip bounds a lot more ofthen than the clip corners and
they are already split up in the shader, so only send the corners if we
need to.
2020-01-07 17:27:15 +01:00
Timm Bäder
2e6e6c1779 gl renderer: Fix color comparison 2020-01-07 17:27:15 +01:00
Timm Bäder
a29b8fbef4 gl renderer: Shorten debug output impl
This will leak a string but WHATEVER.
2020-01-07 17:27:15 +01:00
Timm Bäder
5b072e716c gl renderer: Fix debug color output 2020-01-07 17:27:15 +01:00
Timm Bäder
49845795d9 gl renderer: Rewort passing rounded rects to shaders 2020-01-07 17:27:15 +01:00
Timm Bäder
ab04c74ec9 gl renderer: Don't copy border outline corner sizes around
We already offset + scale the outline and its corners, just pass those
directly to the shader.
2020-01-07 17:27:15 +01:00
Timm Bäder
4744bb9099 gl renderer: stop copying current color around
Just use a pointer now.
2020-01-07 17:27:15 +01:00
Timm Bäder
91522dda63 gl renderer: remove a memory leak 2020-01-07 17:27:15 +01:00
Timm Bäder
f85448ffbf gl renderer: Split blurring a node into its own function 2020-01-07 17:27:15 +01:00
Timm Bäder
d3852ca33a gl renderer: Remove some unused uniforms 2020-01-07 17:27:15 +01:00
Timm Bäder
9d9a730659 gl renderer: Drop a dead if statement
We already handle this earlier.
2020-01-07 17:27:15 +01:00
Timm Bäder
d868b23c76 gl renderer: Implement blurred inset shadow nodes
And with this...

Fixes #1101
2019-12-03 08:40:35 +01:00
Timm Bäder
b91913d10d opbuffer: Remove unused op member 2019-12-03 08:40:35 +01:00
Timm Bäder
1854f3f49c gl renderer: Cache current scale in RenderOpBuilder
We call ops_get_scale a lot, so this should be better. It will also make
a later x/y split for the scaling easier.
2019-12-03 08:40:34 +01:00
Timm Bäder
ca66e9788c gl renderer: Save one level of indentation 2019-12-03 08:40:34 +01:00
Timm Bäder
fb5dc73841 gl renderer: Implement blurred shadow nodes 2019-12-03 08:40:34 +01:00
Timm Bäder
8d61d9f627 gl renderer: Take a TextureRegion in blur_texture
So we can refer to textures on a texture atlas.
2019-12-03 08:40:34 +01:00
Timm Bäder
22b6085c32 gl renderer: Scale blur radius
We scale the texture size, so as a consequence we have to scale the blur
radius.
2019-12-03 08:40:34 +01:00
Timm Bäder
2deb1a05a0 gl renderer: Add some often needed debug code 2019-12-03 08:40:34 +01:00
Timm Bäder
8ca46169b1 gl renderer: Blur shader improvements
Use a two-pass blur shader, fix a few other things and unify the
blurring of blur nodes and blurred outset shadow nodes.

Related to #1283
2019-12-03 08:40:34 +01:00
Timm Bäder
982890515d gl renderer: Don't use g_assert() for code that should always run
g_assert can be compiled out.
2019-11-27 16:56:47 +01:00
Timm Bäder
1817025f46 gl renderer: Don't save repeat nodes to disk
This is just debuggin code someone forgot.
2019-11-27 16:47:14 +01:00
Timm Bäder
8f6de1f955 gl renderer: Load flipped-y offscreen vertex data directly
Instead of loading the unflipped version first and then flipping it.
Don't do it in add_render_ops either but only in the function actually
adding the render ops for the nodes, since those frequently have
early-out conditions that don't need the vertex data at all.
2019-11-26 09:09:25 +01:00
Timm Bäder
91f7ac2d58 gl renderer: Care more about 0×0 offscreen nodes
These never result in a rendering of course, but we must make sure not
to create a 0×0 texture for them, since that will cause problems later
on.
2019-11-26 09:09:25 +01:00
Timm Bäder
053360df85 gl renderer: Remove unused function 2019-11-19 11:06:08 +01:00
Timm Bäder
600ce68210 gl renderer: Add a short cut for repeat nodes that don't repeat
These happen and we don't even need to draw the child to a texture.
2019-11-19 10:45:02 +01:00
Timm Bäder
604f44da11 gl renderer: Fix repeat nodes that don't repeat
Fixes #2234
2019-11-19 10:31:21 +01:00
Matthias Clasen
d4c97ea2b4 gl: Use the fallback debug flag
This debug flag was unused; use it to enable
fallback highlighting at runtime.
2019-10-30 22:31:47 -04:00
Matthias Clasen
5b508ea94a Revert "gl: Speed up icon cache lookups"
This reverts commit dd5ee87b5b.
2019-10-22 07:16:41 -04:00
Matthias Clasen
472d8eebbe gl: Avoid pointless iteration 2019-10-22 07:16:41 -04:00
Matthias Clasen
dd316c8051 gl: Add some comments 2019-10-22 07:16:41 -04:00
Matthias Clasen
1038bc781a Revert "Cache glyph textures in render nodes"
This reverts commit c5af463843.
2019-10-22 07:16:41 -04:00
Matthias Clasen
d3431f569c Revert "inspector: Fix node recording"
This reverts commit ba7649b388.
2019-10-22 01:32:51 -04:00
Matthias Clasen
ba7649b388 inspector: Fix node recording
When attaching renderer-specific data, we need to
make sure that we key it off the renderer that is
in use, and cope with the absence of render data.

This fixes recording nodes in the inspector.
2019-10-18 09:33:45 -05:00
Matthias Clasen
fdbb925654 gl: Remove an unimplemented profiler counter 2019-10-17 07:59:34 -05:00
Matthias Clasen
aeabe3c40e gl: Add debug spew to texture atlas 2019-10-17 07:59:34 -05:00
Matthias Clasen
222b6c2b58 glyph cache: Go back to memcmp
Be careful to avoid padding data, and only
compare the relevant parts, leaving out the
hash key.
2019-10-17 07:59:34 -05:00
Matthias Clasen
5ab5ff7677 Cosmetics 2019-10-17 07:59:34 -05:00
Matthias Clasen
6c92b824f3 Merge branch 'wip/chergert/opbuffer' into 'master'
Add OpBuffer helper for building op buffer

See merge request GNOME/gtk!1131
2019-10-17 11:37:07 +00:00
Matthias Clasen
d777300d4e Fix a crash with glyph caching
We need to treat atlas-less cached glyphs like
atlases, when it comes to invalidating text node
render data.
2019-10-15 22:52:28 -04:00
Matthias Clasen
dd5ee87b5b gl: Speed up icon cache lookups
Use gdk_texture_set_render_data to avoid
hash table lookups when we can.
2019-10-15 19:44:26 -04:00
Matthias Clasen
e34d1b8a26 gl: Slightly rework the icon cache api
Return a pointer to the IconData struct. This is
closer to the glyph cache api, and will allow us
to add similar shortcuts. For now, just store
texture coords in the form we need, avoiding
converting them over and over.
2019-10-15 19:44:26 -04:00
Matthias Clasen
c5af463843 Cache glyph textures in render nodes
This is a quick implementation that avoids many
glyph cache lookups. We keep an array of direct
pointers in the text render node, and throw those
cached pointers away whenever any atlases have
been dropped (since that may invalidate the cached
glyphs).
2019-10-15 19:44:26 -04:00
Matthias Clasen
49748c9c23 Some more vertex data reshuffling
In some cases, the vertex data is just a trivial
modification of the default data, so do that instead
of recalculating it.
2019-10-15 19:44:26 -04:00
Matthias Clasen
b53fa48794 Fix load_vertex_data
There was a copy-paste error that set all
uv coordinates to 0,0.
2019-10-15 19:44:26 -04:00
Christian Hergert
97f3371709 gl: avoid calculating vertex_data until necessary
In many cases of the switch, we do not need the vertex data. This moves
the creation of the vertex_data array into a secondary function and only
calculates it the cases for which it is required.
2019-10-15 19:44:26 -04:00
Matthias Clasen
58d57e1087 gl: Drop buffer_size
No need to maintain buffer_size separately.
It is always vertices->len * sizeof (GskQuadVertex).
2019-10-15 19:44:26 -04:00
Matthias Clasen
571068af12 Drop OP_CHANGE_VAO
Instead, we accumulate vertices in a separate
array, which simplifies various things and lets
us avoid the extra copying step for the vao.
2019-10-15 19:44:26 -04:00
Christian Hergert
a00d12c62a prototype OpBuffer helper for building op buffer 2019-10-15 19:44:26 -04:00
Timm Bäder
5ea21f7910 gl renderer: Fix an out of bounds read
Fixes #2200
2019-10-15 07:13:14 +02:00
Matthias Clasen
7bff3abe8e glyph cache: Fix handling of big glyphs
We were putting big glyphs in the cache, in their
own texture, but forgetting to mark the texture
as permanent, so it could be reused, leading to
occasional misrendering. Fix this by marking these
textures as permanent, and explicitly freeing them
when the cache entry gets old.
2019-10-12 18:53:22 -04:00
Matthias Clasen
e46a7ca706 shadow cache: Remove outdated comments
No comments are better than outdated comments.
2019-10-12 17:06:39 -04:00
Matthias Clasen
e9ba7eda47 gl: Increate the cache check frequency
Otherwise, we spread the cache over more atlases
than necessary, increasing the amount of texture
changes in each frame.
2019-10-12 12:37:11 -04:00
Matthias Clasen
88649b6aae gl: Interleave cache aging
Every few frames, we do extra work for the
cache aging. Arrange for the glyph and icon
caches to not cause extra work on the same
frame, to smooth things out.
2019-10-12 12:37:11 -04:00
Matthias Clasen
60d63bbada gl: Improve debug spew for caches
Dump similar information for both caches,
and correct the unused percentage for
the atlases.
2019-10-12 12:37:11 -04:00
Matthias Clasen
123cbd42bb gl: Make icon cache work like glyph cache
Replace timestamp tracking with an accessed bit
here too, to keep the glyph and icon cache code
similar.
2019-10-12 11:46:10 -04:00
Matthias Clasen
977ac2b31f gl: Do less work on glyph caching
There is no need for us to be very precise about
aging the glyph entries. It is enough to check
occasionally and mark old entries. This reduces
the overhead of work we do every frame on the
caches, at the cost of letting glyphs linger
a bit longer in the cache.
2019-10-12 11:35:46 -04:00
Timm Bäder
bcdc3b706c iconcache: Fix icon padding 2019-10-12 09:17:24 +02:00
Matthias Clasen
61db797f29 gl: Simplify glyph cache lookup
Make this function more similar to the icon
cache equivalent, and simplify it a bit. We
don't use the boolean return, and we don't need
to look at the age of entry when marking it
used.
2019-10-12 01:35:13 -04:00
Matthias Clasen
0a876f11a0 gl: Don't use memcmp for comparing cache keys
Some innocent change made us use a stack-allocated
key, and things broke. Lets go back to comparing
cache keys field by field.
2019-10-12 01:00:08 -04:00
Matthias Clasen
1c17316f9c gl: Handle row stride for icon cache upload
Same as the previous commit: Downloading a texture
may in theory give us data with a stride, so handle
that.
2019-10-11 21:17:30 -04:00
Matthias Clasen
646c5f369f gl: Handle row stride for glyph cache upload
In theory, we can have data with a stride here,
so set the necessary parameters to tell GL about
it.
2019-10-11 21:16:40 -04:00
Matthias Clasen
8839e10d44 gl: Do less work to maintain caches
Remember which atlases were removed, and only
check those when looking for icons or glyphs
to remove. For most frames, we don't have to
check at all since no atlases were removed.
2019-10-11 20:42:24 -04:00