Previously Cogl would only ever use one atlas for textures and if it
reached the maximum texture size then all other new textures would get
their own GL texture. This patch makes it so that we create as many
atlases as needed. This should avoid breaking up some batches and it
will be particularly good if we switch to always using multi-texturing
with a default shader that selects between multiple atlases using a
vertex attribute.
Whenever a new atlas is created it is stored in a GSList on the
context. A weak weference is taken on the atlas using
cogl_object_set_user_data so that it can be removed from the list when
the atlas is destroyed. The atlas textures themselves take a reference
to the atlas and this is the only thing that keeps the atlas
alive. This means that once the atlas becomes empty it will
automatically be destroyed.
All of the COGL_NOTEs pertaining to atlases are now prefixed with the
atlas pointer to make it clearer which atlas is changing.
Once the GLES2 wrapper is removed we won't be able to upload the
matrices with the fixed function API any more. The fixed function API
gives a global state for setting the matrix but if a custom shader
uniform is used for the matrices then the state is per
program. _cogl_matrix_stack_flush_to_gl is called in a few places and
it is assumed the current pipeline doesn't need to be flushed before
it is called. To allow these semantics to continue to work, on GLES2
the matrix flush now just stores a reference to the matrix stack in
the CoglContext. A pre_paint virtual is added to the progend which is
called whenever a pipeline is flushed, even if the same pipeline was
flushed already. This gives the GLSL progend a chance to upload the
matrices to the uniforms. The combined modelview/projection matrix is
only calculated if it is used. The generated programs end up never
using the modelview or projection matrix so it usually only has to
upload the combined matrix. When a matrix stack is flushed a reference
is taked to it by the pipeline progend and the age is stored so that
if the same state is used with the same program again then we don't
need to reupload the uniform.
When the GLES2 wrapper is removed we can't use the fixed function API
such as glColorPointer to set the builtin attributes. Instead the GLSL
progend now maintains a cache of attribute locations that are queried
with glGetAttribLocation. The code that previously maintained a cache
of the enabled texture coord arrays has been modified to also cache
the enabled vertex attributes under GLES2. The vertex attribute API is
now the only place that is using this cache so it has been moved into
cogl-vertex-attribute.c
The GLSL vertend is mostly only useful for GLES2. The fixed function
vertend is kept at higher priority than the GLSL vertend so it is
unlikely to be used in any other circumstances.
We were trying to declare and initializing an arbfp program cache for
GLES but since the prototypes for the _hash and _equal functions were
only available for GL this broke the GLES builds. By #ifdefing the code
to conditionally declare/initialize for GL only this should hopefully
fix GLES builds.
This adds a cache (A GHashTable) of ARBfp programs and before ever
starting to code-generate a new program we will always first try and
find an existing program in the cache. This uses _cogl_pipeline_hash and
_cogl_pipeline_equal to hash and compare the keys for the cache.
There is a new COGL_DEBUG=disable-program-caches option that can disable
the cache for debugging purposes.
Before flushing the journal there is now a separate iteration that
will try to determine if the matrix of the clip stack and the matrix
of the rectangle in each entry are on the same plane. If they are it
can completely avoid the clip stack and instead manually modify the
vertex and texture coordinates to implement the clip. The has the
advantage that it won't break up batching if a single clipped
rectangle is used in a scene.
The software clip is only used if there is no user program and no
texture matrices. There is a threshold to the size of the batch where
it is assumed that it is worth the cost to break up a batch and
program the GPU to do the clipping. Currently this is set to 8
although this figure is plucked out of thin air.
To check whether the two matrices are on the same plane it tries to
determine if one of the matrices is just a simple translation of the
other. In the process of this it also works out what the translation
would be. These values can be used to translate the clip rectangle
into the coordinate space of the rectangle to be logged. Then we can
do the clip directly in the rectangle's coordinate space.
When logging a quad we now only store the 2 vertices representing the
top left and bottom right of the quad. The color is only stored once
per entry. Once we come to upload the data we expand the 2 vertices
into four and copy the color to each vertex. We do this by mapping the
buffer and directly expanding into it. We have to copy the data before
we can render it anyway so it doesn't make much sense to expand the
vertices before uploading and this way should save some space in the
size of the journal. It also makes it slightly easier if we later want
to do pre-processing on the journal entries before uploading such as
doing software clipping.
The modelview matrix is now always copied to the journal entry whereas
before it would only be copied if we aren't doing software
transform. The journal entry struct always has the space for the
modelview matrix so hopefully it's only a small cost to copy the
matrix.
The transform for the four entries is now done using
cogl_matrix_transform_points which may be slightly faster than
transforming them each individually with a call to
cogl_matrix_transfom.
The features_cached member of CoglContext is intended to mark when
we've calculated the features so that we know if they are ready in
cogl_get_features. However we always intialize the features while
creating the context so features_cached will never be FALSE so it's
not useful. We also had the odd behaviour that the COGL_DEBUG feature
overrides were only applied in the first call to
cogl_get_features. However there are other functions that use the
feature flags such as cogl_features_available that don't use this
function so in some cases the feature flags will be interpreted before
the overrides are applied. This patch makes it always initialize the
features and apply the overrides immediately while creating the
context. This fixes a problem with COGL_DEBUG=disable-arbfp where the
first material flushed is done before any call to cogl_get_features so
it may still use ARBfp.
The GLSL pipeline backend can now generate code to represent the
pipeline state in a similar way to the ARBfp backend. Most of the code
for this is taken from the GLES 2 wrapper.
Only one of the material backends can be generating code at the same
time so it seems to make sense to share the same source buffer between
arbfp and glsl. The new name is fragment_source_buffer in case we
later want to create a new buffer for the vertex shader. That probably
couldn't share the same buffer because it will likely need to be
generated at the same time.
When COGL_ENABLE_EXPERIMENTAL_2_0_API is defined cogl.h will now include
cogl2-path.h which changes cogl_path_new() so it can directly return a
CoglPath pointer; it no longer exposes a prototype for
cogl_{get,set}_path and all the remaining cogl_path_ functions now take
an explicit path as their first argument.
The idea is that we want to encourage developers to retain path objects
for as long as possible so they can take advantage of us uploading the
path geometry to the GPU. Currently although it is possible to start a
new path and query the current path, it is not convenient.
The other thing is that we want to get Cogl to the point where nothing
depends on a global, current context variable. This will allow us to one
day define a sensible threading model if/when that is ever desired.
When using cogl_set_source_color4ub there is a notable difference
between colors that require blending and those that dont. When trying to
modify the color of pipeline referenced by the journal we don't force a
flush of the journal unless the color change will also change the
blending state. By using two separate pipeline objects for handing
opaque or transparent colors we can avoid ever flushing the journal when
repeatedly using cogl_set_source_color and jumping between opaque and
transparent colors.
Previously we tracked whether the clip stack needs flushing as part of
the CoglClipState which is part of the CoglFramebuffer state. This is
a bit odd because most of the clipping state (such as the clip planes
and the scissor) are part of the GL context's state rather than the
framebuffer. We were marking the clip state on the framebuffer dirty
every time we change the framebuffer anyway so it seems to make more
sense to have the dirtiness be part of the global context.
Instead of a just a single boolean to record whether the state needs
flushing, the CoglContext now holds a reference to the clip stack that
was flushed. That way we can flush arbitrary stack states and if it
happens to be the same as the state already flushed then Cogl will do
nothing. This will be useful if we log the clip stack in the journal
because then we will need to flush unrelated clip stack states for
each batch.
This applies an API naming change that's been deliberated over for a
while now which is to rename CoglMaterial to CoglPipeline.
For now the new pipeline API is marked as experimental and public
headers continue to talk about materials not pipelines. The CoglMaterial
API is now maintained in terms of the cogl_pipeline API internally.
Currently this API is targeting Cogl 2.0 so we will have time to
integrate it properly with other upcoming Cogl 2.0 work.
The basic reasons for the rename are:
- That the term "material" implies to many people that they are
constrained to fragment processing; perhaps as some kind of high-level
texture abstraction.
- In Clutter they get exposed by ClutterTexture actors which may be
re-inforcing this misconception.
- When comparing how other frameworks use the term material, a material
sometimes describes a multi-pass fragment processing technique which
isn't the case in Cogl.
- In code, "CoglPipeline" will hopefully be a much more self documenting
summary of what these objects represent; a full GPU pipeline
configuration including, for example, vertex processing, fragment
processing and blending.
- When considering the API documentation story, at some point we need a
document introducing developers to how the "GPU pipeline" works so it
should become intuitive that CoglPipeline maps back to that
description of the GPU pipeline.
- This is consistent in terminology and concept to OpenGL 4's new
pipeline object which is a container for program objects.
Note: The cogl-material.[ch] files have been renamed to
cogl-material-compat.[ch] because otherwise git doesn't seem to treat
the change as a moving the old cogl-material.c->cogl-pipeline.c and so
we loose all our git-blame history.
This updates the implementation of cogl_polygon so it sits on the new
CoglVertexArray and CoglVertexAttribute apis. This lets us minimize the
number of different drawing paths we have to maintain in Cogl.
Since the sliced texture support for cogl_polygon has been broken for a
long time now and no one has complained this patch also greatly
simplifies the code by not doing any special material validation so
cogl_polygon will be restricted in the same way as
cogl_draw_vertex_attributes. (i.e. sliced textures not supported).
Instead of using raw OpenGL in the journal we now use the vertex
attributes API instead. This is part of an ongoing effort to reduce the
number of drawing paths we maintain in Cogl.
The functionality of cogl_vertex_buffer_indices_get_for_quads is now
provided by cogl_get_rectangle_indices so this reworks the former to now
work in terms of the latter so we don't have duplicated logic.
CoglIndices define a range of indices inside a CoglIndexArray. I.e. a
CoglIndexArray is simply a buffer of N bytes and you can then
instantiate multiple CoglIndices collections that define a sub-region of
a CoglIndexArray by specifying a start offset and an index data type.
This allows _cogl_material_flush_gl_state to bail out faster if
repeatedly asked to flush the same material and we can see the material
hasn't changed.
Since we can rely on the material age incrementing when any material
property changes or any associated layer property changes then we can
track the age of the material after flushing so it can be compared with
the age of the material if it is subsequently re-flushed. If the age is
the same we only have to re-assert the texture object state.
Previously cogl_set_source_color and cogl_set_source_texture modified
a single global material. If an application then mixes using
cogl_set_source_color and texture then the material will constantly
need a new ARBfp program because the numbers of layers alternates
between 0 and 1. This patch just adds a second global material that is
only used for cogl_set_source_texture. I think it would still end up
flushing the journal if cogl_set_source_texture is used with multiple
different textures but at least it should avoid a recompile unless the
texture target also changes. It might be nice to somehow attach a
material to the CoglTexture for use with cogl_set_source_texture but
it would be difficult to implement this without creating a circular
reference.
This exposes the idea of a stack of source materials instead of just
having a single current material. This allows the writing of orthogonal
code that can change the current source material and restore it to its
previous state. It also allows the implementation of new composite
primitives that may want to validate the current source material and
possibly make override changes in a derived material.
*** WARNING: THIS COMMIT CHANGES THE BUILD ***
Do not recurse into the backend directories to build private, internal
libraries.
We only recurse from clutter/ into the cogl sub-directory; from there,
we don't recurse any further. All the backend-specific code in Cogl and
Clutter is compiled conditionally depending on the macros defined by the
configure script.
We still recurse from the top-level directory into doc, clutter and
tests, because gtk-doc and tests do not deal nicely with non-recursive
layouts.
This change makes Clutter compile slightly faster, and cleans up the
build system, especially when dealing with introspection data.
Ideally, we also want to make Cogl part of the top-level build, so that
we can finally drop the sed trick to change the shared library from the
GIR before compiling it.
Currently disabled:
‣ OSX backend
‣ Fruity backend
Currently enabled but untested:
‣ EGL backend
‣ Windows backend
Instead of storing a pointer to the CoglRectangleMap and a handle to
the atlas texture in the context, there is a now a separate data
structure called a CoglAtlas to manage these two. The context just
contains a pointer to this. The code to reorganise the atlas has been
moved from cogl-atlas-texture.c to cogl-atlas.c
This simply renames CoglAtlas to CoglRectangleMap without making any
functional changes. The old 'CoglAtlas' is just a data structure for
managing unused areas of a rectangle and it doesn't neccessarily have
to be used for an atlas so it wasn't a very good name.
Previously cogl_set_fog would cause a flush of the Cogl journal and
would directly bang the GL state machine to setup fogging. As part of
the ongoing effort to track most state in CoglMaterial to support
renderlists this now adds an indirection so that cogl_set_fog now just
updates ctx->legacy_fog_state. The fogging state then gets enabled as a
legacy override similar to how the old depth testing API is handled.
This makes CoglBuffer track the last used bind target as a private
property. This is later used when binding a buffer to map instead of
always using the PIXEL_UNPACK target.
This also adds some additional sanity checks that code doesn't try to
nest binds to the same target or bind a buffer to multiple targets at
the same time.
Instead of having to extend cogl_is_buffer with new buffer types
manually this now adds a new COGL_BUFFER_DEFINE macro to be used instead
of COGL_OBJECT_DEFINE for CoglBuffer subclasses. This macro will
automatically register the new type with ctx->buffer_types which will
iterated by cogl_is_buffer. This is the same coding pattern used for
CoglTexture.
As part of the ongoing effort to remove CoglHandle from the API this
switches the cogl_material API to use a strongly typed CoglMaterial
pointer instead of CoglHandle.
This creates a separate struct to store the fields of the context that
are specific to the winsys. This is all stored in one file but ideally
this could work more like the CoglContextDriver struct and have a
different header for each winsys.
Instead of having a hardcoded series of if-statements in
cogl_is_texture to determine which types should appear as texture
subclasses, they are now stored in a GSList attached to the Cogl
context. The list is amended to using a new cogl_texture_register_type
function. There is a convenience macro called COGL_TEXTURE_DEFINE
which uses COGL_HANDLE_DEFINE_WITH_CODE to register the texture type
when the _get_type() function is first called.
This redirects the legacy depth testing APIs through CoglMaterial and
adds a new experimental cogl_material_ API for handling the depth
testing state.
This adds the following new functions:
cogl_material_set_depth_test_enabled
cogl_material_get_depth_test_enabled
cogl_material_set_depth_writing_enabled
cogl_material_get_depth_writing_enabled
cogl_material_set_depth_test_function
cogl_material_get_depth_test_function
cogl_material_set_depth_range
cogl_material_get_depth_range
As with other experimental Cogl API you need to define
COGL_ENABLE_EXPERIMENTAL_API to access them and their stability isn't
yet guaranteed.
This is a complete overhaul of the data structures used to manage
CoglMaterial state.
We have these requirements that were aiming to meet:
(Note: the references to "renderlists" correspond to the effort to
support scenegraph level shuffling of Clutter actor primitives so we can
minimize GPU state changes)
Sparse State:
We wanted a design that allows sparse descriptions of state so it scales
well as we make CoglMaterial responsible for more and more state. It
needs to scale well in terms of memory usage and the cost of operations
we need to apply to materials such as comparing, copying and flushing
their state. I.e. we would rather have these things scale by the number
of real changes a material represents not by how much overall state
CoglMaterial becomes responsible for.
Cheap Copies:
As we add support for renderlists in Clutter we will need to be able to
get an immutable handle for a given material's current state so that we
can retain a record of a primitive with its associated material without
worrying that changes to the original material will invalidate that
record.
No more flush override options:
We want to get rid of the flush overrides mechanism we currently use to
deal with texture fallbacks, wrap mode changes and to handle the use of
highlevel CoglTextures that need to be resolved into lowlevel textures
before flushing the material state.
The flush options structure has been expanding in size and the structure
is logged with every journal entry so it is not an approach that scales
well at all. It also makes flushing material state that much more
complex.
Weak Materials:
Again for renderlists we need a way to create materials derived from
other materials but without the strict requirement that modifications to
the original material wont affect the derived ("weak") material. The
only requirement is that its possible to later check if the original
material has been changed.
A summary of the new design:
A CoglMaterial now basically represents a diff against its parent.
Each material has a single parent and a mask of state that it changes.
Each group of state (such as the blending state) has an "authority"
which is found by walking up from a given material through its ancestors
checking the difference mask until a match for that group is found.
There is only one root node to the graph of all materials, which is the
default material first created when Cogl is being initialized.
All the groups of state are divided into two types, such that
infrequently changed state belongs in a separate "BigState" structure
that is only allocated and attached to a material when necessary.
CoglMaterialLayers are another sparse structure. Like CoglMaterials they
represent a diff against their parent and all the layers are part of
another graph with the "default_layer_0" layer being the root node that
Cogl creates during initialization.
Copying a material is now basically just a case of slice allocating a
CoglMaterial, setting the parent to be the source being copied and
zeroing the mask of changes.
Flush overrides should now be handled by simply relying on the cheapness
of copying a material and making changes to it. (This will be done in a
follow on commit)
Weak material support will be added in a follow on commit.
As part of an effort to improve the architecture of CoglMaterial
internally this overhauls how we flush layer state to OpenGL by adding a
formal backend abstraction for fragment processing and further
formalizing the CoglTextureUnit abstraction.
There are three backends: "glsl", "arbfp" and "fixed". The fixed backend
uses the OpenGL fixed function APIs to setup the fragment processing,
the arbfp backend uses code generation to handle fragment processing
using an ARBfp program, and the GLSL backend is currently only there as
a formality to handle user programs associated with a material. (i.e.
the glsl backend doesn't yet support code generation)
The GLSL backend has highest precedence, then arbfp and finally the
fixed. If a backend can't support some particular CoglMaterial feature
then it will fallback to the next backend.
This adds three new COGL_DEBUG options:
* "disable-texturing" as expected should disable all texturing
* "disable-arbfp" always make the arbfp backend fallback
* "disable-glsl" always make the glsl backend fallback
* "show-source" show code generated by the arbfp/glsl backends
The Cogl context has now a feature_flags_private enum that will allow us
to query and use OpenGL features without exposing them in the public
API.
The ARB_fragment_program extension is the first user of those flags.
Looking for this extension only happens in the gl driver as the gles
drivers will not expose them.
One can use _cogl_features_available_private() to check for the
availability of such private features.
While at it, reindent cogl-internal.h as described in CODING_STYLE.
Instead of directly using a guint32 to store a bitmask for each used
texcoord array, it now stores them in a CoglBitmask. This removes the
limitation of 32 layers (although there are still other places in Cogl
that imply this restriction). To disable texcoord arrays code should
call _cogl_disable_other_texcoord_arrays which takes a bitmask of
texcoord arrays that should not be disabled. There are two extra
bitmasks stored in the CoglContext which are used temporarily for this
function to avoid allocating a new bitmask each time.
http://bugzilla.openedhand.com/show_bug.cgi?id=2132
It should be quite acceptable to use a texture without defining any
texture coords. For example a shader may be in use that is doing
texture lookups without referencing the texture coordinates. Also it
should be possible to replace the vertex colors using a texture layer
without a texture but with a constant layer color.
enable_state_for_drawing_buffer no longer sets any disabled layers in
the overrides. Instead of counting the number of units with texture
coordinates it now keeps them in a mask. This means there can now be
gaps in the list of enabled texture coordinate arrays. To cope with
this, the Cogl context now also stores a mask to track the enabled
arrays. Instead of code manually iterating each enabled array to
disable them, there is now an internal function called
_cogl_disable_texcoord_arrays which disables a given mask.
I think this could also fix potential bugs when a vertex buffer has
gaps in the texture coordinate attributes that it provides. For
example if the vertex buffer only had texture coordinates for layer 2
then the disabling code would not disable the coordinates for layers 0
and 1 even though they are not used. This could cause a crash if the
previous data for those arrays is no longer valid.
http://bugzilla.openedhand.com/show_bug.cgi?id=2132
This adds three new API calls:
CoglHandle cogl_path_get()
void cogl_path_set(CoglHandle path)
CoglHandle cogl_path_copy(CoglHandle path)
All of the fields relating to the path have been moved from the Cogl
context to a new CoglPath handle type. The cogl context now just
contains a CoglPath handle. All of the existing path commands
manipulate the data in the current path handle. cogl_path_new now just
creates a new path handle and unrefs the old one.
The path handle can be stored for later with cogl_path_get. The path
can then be copied with cogl_path_copy. Internally it implements
copy-on-write semantics with an extra optimisation that it will only
copy the data if the new path is modified, but not if the original
path is modified. It can do this because the only way to modify a path
is by appending to it so the copied path is able to store its own path
length and only render the nodes up to that length. For this to work
the copied path also needs to keep its own copies of the path extents
because the parent path may change these by adding nodes.
The clip stack now uses the cogl_path_copy mechanism to store paths in
the stack instead of directly copying the data. This should save some
memory and processing time.
Since using addresses that might change is something that finally
the FSF acknowledge as a plausible scenario (after changing address
twice), the license blurb in the source files should use the URI
for getting the license in case the library did not come with it.
Not that URIs cannot possibly change, but at least it's easier to
set up a redirection at the same place.
As a side note: this commit closes the oldes bug in Clutter's bug
report tool.
http://bugzilla.openedhand.com/show_bug.cgi?id=521
The function _cogl_get_max_texture_units is called quite often while
rendering and it returns a constant value so we might as well cache
the result. Calling glGetInteger on Mesa can be expensive because it
flushes a lot of state.
We've had complaints that our Cogl code/headers are a bit "special" so
this is a first pass at tidying things up by giving them some
consistency. These changes are all consistent with how new code in Cogl
is being written, but the style isn't consistently applied across all
code yet.
There are two parts to this patch; but since each one required a large
amount of effort to maintain tidy indenting it made sense to combine the
changes to reduce the time spent re indenting the same lines.
The first change is to use a consistent style for declaring function
prototypes in headers. Cogl headers now consistently use this style for
prototypes:
return_type
cogl_function_name (CoglType arg0,
CoglType arg1);
Not everyone likes this style, but it seems that most of the currently
active Cogl developers agree on it.
The second change is to constrain the use of redundant glib data types
in Cogl. Uses of gint, guint, gfloat, glong, gulong and gchar have all
been replaced with int, unsigned int, float, long, unsigned long and char
respectively. When talking about pixel data; use of guchar has been
replaced with guint8, otherwise unsigned char can be used.
The glib types that we continue to use for portability are gboolean,
gint{8,16,32,64}, guint{8,16,32,64} and gsize.
The general intention is that Cogl should look palatable to the widest
range of C programmers including those outside the Gnome community so
- especially for the public API - we want to minimize the number of
foreign looking typedefs.
Buffer objects are cool! This abstracts the buffer API first introduced
by GL_ARB_vertex_buffer_object and then extended to other objects.
The coglBuffer abstract class is intended to be the base class of all
the buffer objects, letting the user map() buffers. If the underlying
implementation does not support buffer objects (or only support VBO but
not FBO for instance), fallback paths should be provided.
The internal format of the atlas texture is still set to the
appropriate format so Cogl will disable blending for textures that are
intended to be RGB. This should end up ignoring the alpha channel from
the texture in the atlas. This makes the code slightly easier to
maintain and should also improve the chances of batching.
Instead of assigning a new colour to each quad of a batch, the
rectangle debugging code now assigns a new colour to each batch so
that it can be used to visually see what is being batched. The colour
is stored in a global variable that is reset during cogl_clear. This
improves the chances that the same colour will be used for a batch in
the next frames to avoid flickering.
This adds a CoglAtlas type which is a data structure that keeps track
of unused sub rectangles of a larger rectangle. There is a new atlased
texture backend which uses this to put multiple textures into a single
larger texture.
Currently the atlas is always sized 256x256 and the textures are never
moved once they are put in. Eventually it needs to be able to
reorganise the atlas and grow it if necessary. It also needs to
migrate the textures out of the atlas if mipmaps are required.
cogl_push_draw_buffer, cogl_set_draw_buffer and cogl_pop_draw_buffer are now
deprecated and new code should use the new cogl_framebuffer_* API instead.
Code that previously did:
cogl_push_draw_buffer ();
cogl_set_draw_buffer (COGL_OFFSCREEN_BUFFER, buffer);
/* draw */
cogl_pop_draw_buffer ();
should now be re-written as:
cogl_push_framebuffer (buffer);
/* draw */
cogl_pop_framebuffer ();
As can be seen from the example above the rename has been used as an
opportunity to remove the redundant target argument from
cogl_set_draw_buffer; it now only takes one call to redirect to an offscreen
buffer, and finally the term framebuffer may be a bit more familiar to
anyone coming from an OpenGL background.
cogl_material_copy can be used to create a new CoglHandle referencing a copy
of some given material.
From now on we will advise that developers always aim to use this function
instead of cogl_material_new() when creating a material that is in any way
derived from another.
By using cogl_material_copy, Cogl can maintain an ancestry for each material
and keep track of "similar" materials. The plan is that Cogl will use this
information to minimize the cost of GPU state transitions.
