Commit Graph

2436 Commits

Author SHA1 Message Date
Robert Bragg
1cc3ae6944 CoglMaterial: Implements sparse materials design
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.
2010-06-15 15:26:27 +01:00
Robert Bragg
6c8b8cef9e material: fix #ifdef guard around use of _TEXTURE_RECTANGLE
We were incorrectly guarding the use of GL_TEXTURE_RECTANGLE_ARB with
ifdef ARB_texture_rectangle instead of ifdef GL_ARB_texture_rectangle
which broke test-cogl-texture-rectangle.
2010-06-15 15:26:27 +01:00
Robert Bragg
26628553ed blend-strings: removes the AUTO_COMPOSITE function enum
This was mistakenly added some time ago because at some point when we
were discussing how to handle premultiplied alpha in Clutter/Cogl we
were considering having a magic "just do the right thing" option which
was later abandoned.
2010-06-15 15:26:27 +01:00
Robert Bragg
99bef3116d cogl-color: add cogl_color_init_from_xyz funcs
This is to try and improve API consistency. Simple cogl structures that
don't derive from CoglObject and which can be allocated on the stack,
such as CoglColor and CoglMatrix should all have  "_init" or
"_init_from" functions to initialize all the structure members. (As
opposed to a cogl_xyz_new() function for CoglObjects). CoglColor
previously used the naming scheme "_set_from" for these initializers but
"_set" is typically reserved for setting individual properties of a
structure/object.

This adds three _init functions:
cogl_color_init_from_4ub
cogl_color_init_from_4f
cogl_color_init_from_4fv

The _set_from functions are now deprecated but only with a gtk-doc
annotation for now. This is because the cogl_color_set_from API is quite
widely used already and so were giving a grace period before enabling a
GCC deprecated warning just because otherwise the MX maintainers will
complain to me that I've made their build logs look messy.
2010-06-15 15:26:27 +01:00
Robert Bragg
49907d6267 cogl-color: add private _cogl_color_get_rgba_4ubv API
The journal logs colors as 4bytes into a vertex array and since we are
planning to make CoglMaterial track its color using a CoglColor instead
of a byte array this convenience will be useful for re-implementing
_cogl_material_get_colorubv.
2010-06-15 15:26:27 +01:00
Emmanuele Bassi
3af1dc18b8 Merge branch 'wip/deform-effect'
* wip/deform-effect:
  docs: Add DeformEffect and PageTurnEffect to the API reference
  effect: Add PageTurnEffect
  effect: Add DeformEffect
  offscreen-effect: Traslate the modelview with the offsets
  docs: Fix Effect subclassing section
2010-06-11 18:12:54 +01:00
Neil Roberts
fcfff3dd04 cogl/winsys/cogl-eglx: Use the right egl header on 1.1
Commit b0cc98fdce introduced a #include for the egl header in
cogl-eglx.c The header name is different under GLES 1.1 so it broke
compilation.
2010-06-11 16:06:31 +01:00
Neil Roberts
cc7722389c cogl-program: Add cogl_program_uniform_1i to GLES
The GLES 1.1 backend was missing a stub for cogl_program_uniform_1i.
2010-06-11 16:06:31 +01:00
Neil Roberts
9343cb849a cogl-gles2: Prefix internal GLES 2 wrapper symbols with _
Some internal symbols used for the GLES 2 wrapper were accidentally
being exported. This prepends an underscore to them so they won't
appear in the shared library.
2010-06-11 16:06:31 +01:00
Neil Roberts
0bbf50f905 cogl-clip-stack: Always use the scissor when clipping
Whenever a path or a rectangle is added to the clip stack it now also
stores a screen space bounding box in the entry. Then when the clip
stack is flushed the bounding box is first used to set up the
scissor. That way when we eventually come to use the stencil buffer
the clear will be affected by the scissor so we don't have to clear
the entire buffer.
2010-06-10 21:52:55 +01:00
Neil Roberts
5a71ff4cdd cogl-path: Export _cogl_path_get_bounds
_cogl_path_get_bounds is no longer static and is exported in
cogl-path-private.h so that it can be used in the clip stack code. The
old version of the function returned x/y and width/height. However
this was mostly used to call cogl_rectangle which takes x1/y1
x2/y2. The function has been changed to just directly return the
second form because it is more useful. Anywhere that was previously
using the function now just directly looks at path->path_nodes_min and
path->path_nodes_max instead.
2010-06-10 21:52:54 +01:00
Neil Roberts
a9da7d72db cogl-clip-state: Export transform_point internally to Cogl
The transform_point function takes a modelview matrix, projection
matrix and a viewport and performs all three transformations on a
point to give a Cogl window coordinate. This is useful in a number of
places in Cogl so this patch moves it to cogl.c and adds it to
cogl-internal.h
2010-06-10 21:52:49 +01:00
Sunil Sadasivan
4abd42aced cogl_texture_get_data() copies using wrong width size.
For sliced 2D textures, _cogl_texture_2d_sliced_get_data() uses the
bitmap width, instead of the rowstride, when memcpy()ing into the
dest buffer.

Signed-off-by: Robert Bragg <robert@linux.intel.com>
2010-06-10 20:58:22 +01:00
Damien Lespiau
98f2d2516b cogl-color: Provide setters for all the channels
We only had getters for the red, green, blue and alpha channels of a
color. This meant that, if you wanted to change, say, the alpha
component of a color, one would need to query the red, green and blue
channels and use set_from_4ub() or set_from_4f().

Instead of this, just provide some setters for CoglColor, using the same
naming scheme than the existing getters.
2010-06-10 20:02:01 +01:00
Damien Lespiau
896dd975d1 cogl-color: Add cogl_color_unpremultiply()
For some operations on pre-multiplied colors (say, replace the alpha
value), you need to unpremultiply the color.

This patch provides the counterpart to cogl_color_premultiply().
2010-06-10 20:01:48 +01:00
Emmanuele Bassi
2ba6ab6dbc cogl: Initialize the framebuffer bits on state flush
The place where we actually change the framebuffer is
_cogl_framebuffer_flush_state(), so if we changed to a new frame buffer
we need to initialize the color bits there.

http://bugzilla.openedhand.com/show_bug.cgi?id=2094
2010-06-10 19:55:18 +01:00
Emmanuele Bassi
7a7dedebd5 cogl-framebuffer: Use the FBO extension for color sizes
OpenGL 3.0 deprecated querying of the GL_{RED,GREEN,BLUE}_BITS
constants, and the FBO extension provides a mechanism to query for the
color buffer sizes which *should* work even with the default
framebuffer. Unfortunately, this doesn't seem to hold for Mesa - so we
just use this for the offscreen CoglFramebuffer type, and we fall back
to glGetIntegerv() for the onscreen one.

http://bugzilla.openedhand.com/show_bug.cgi?id=2094
2010-06-10 19:53:39 +01:00
Emmanuele Bassi
3c4451cd84 cogl: Hide cogl_framebuffer_get_*_bits()
These accessors should be private, for the time being.

http://bugzilla.openedhand.com/show_bug.cgi?id=2094
2010-06-10 19:06:37 +01:00
Neil Roberts
cce5616edf cogl-material: Fix the function which sets the enable blend flag
This function had two problems. Firstly it would clear the enable
blend flag before calling pre_change_notify so that if blending was
previously enabled the journal would end up being flushed while the
flag was still cleared. Secondly it would call the pre change notify
whenever blending is needed regardless of whether it was already
needed previously.

This was causing problems in test-depth.
2010-06-09 17:26:15 +01:00
Robert Bragg
82e80e6765 material: Avoid redundant glBindTexture calls
This adds a _cogl_bind_gl_texture_transient function that should be used
instead of glBindTexture so we can have a consistent cache of the
textures bound to each texture unit so we can avoid some redundant
binding.
2010-06-09 17:26:15 +01:00
Robert Bragg
acc44161c1 material: Adds backend abstraction for fragment processing
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
2010-06-09 17:15:59 +01:00
Neil Roberts
8ebf76a9a9 cogl-atlas-texture: Flush the journal before starting blit not after
_cogl_atlas_texture_blit_begin binds a texture to use as the
destination and it expects it to stay bound until
_cogl_atlas_texture_end_blit is called. However there was a call to
_cogl_journal_flush directly after setting up the blit state which
could cause the wrong texture to be bound. This just moves the flush
to before the call to _cogl_atlas_texture_blit_begin.

This was breaking test-cogl-sub-texture.
2010-06-09 15:22:05 +01:00
Robert Bragg
4ac3133208 atlas: fix some journal flushing issues
1) Always flush when migrating textures out of an atlas because although
it's true that the original texture data will remain valid in the
original texture we can't assume that journal entries have resolved the
GL texture that will be used. This is only true if a layer0_override has
been used.

2) Don't flush at the point of creating a new atlas simply flush
immediately before reorganizing an atlas. This means we are now assuming
that we will never see recursion due to atlas textures being modified
during a journal flush. This means it's the responsibility of the
primitives code to _ensure_mipmaps for example not the responsibility of
_cogl_material_flush_gl_state.
2010-06-09 15:19:31 +01:00
Robert Bragg
a21d909129 primitives: make sure to update texture storage before drawing
We want to make sure that the material state flushing code will never
result in changes to the texture storage for that material. So for
example mipmaps need to be ensured by the primitives code.

Changes to the texture storage will invalidate the texture coordinates
in the journal and we want to avoid a recursion of journal flushing.
2010-06-09 15:19:31 +01:00
Robert Bragg
1047c1c082 matrix: add cogl_matrix_equal API
This adds a way to compare two CoglMatrix structures to see if they
represent the same transformations. memcmp can't be used because a
CoglMatrix contains private flags and padding.
2010-06-09 15:19:31 +01:00
Damien Lespiau
8ee099d204 cogl-material: Use fragment programs instead of texture env combiners
THIS IS A WORK IN PROGRESS

Mesa is building a big shader when using ARB_texture_env_combine. The
idea is to bypass that computation, do it ourselves and cache the
compiled program in a CoglMaterial.

For now that feature can be enabled by setting the COGL_PIPELINE
environment variable to "arbfp". COGL_SHOW_FP_SOURCE can be set to a non
empty string to dump the fragment program source too.

TODO:
  * fog (really easy, using OPTION)
  * support tex env combiner operands, DOT3, ADD_SIGNED, INTERPOLATE
    combine modes (need refactoring the generation of temporary
    variables) (not too hard)
  * alpha testing for GLES 2.0?
2010-06-09 15:19:30 +01:00
Damien Lespiau
56dd71dba0 cogl: Introduce private feature flags and check for ARB_fp
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.
2010-06-09 15:19:30 +01:00
Neil Roberts
8836c868a4 cogl-gles2-wrapper: Fix some cases using the wrong texture unit
At two places in cogl_wrap_prepare_for_draw it was trying to loop over
the texture units to flush some state. However it was retrieving the
texture unit pointer using w->active_texture_unit instead of the loop
index so it would end up with the wrong state.

Also in glEnableClientState it was using the active unit instead of
the client active unit.
2010-06-08 17:26:26 +01:00
Damien Lespiau
2e52c5f67b doc: Some fixes for the Cogl documentation
• 3 general fixes (typos, copy/paste),
 • ignore cogl-object-private.h,
 • cogl_fixed_atani() was in reality cogl_fixed_atan(), fixed in commit
   43564f05.
 • Fix the cogl-vector section: sections must have a </SECTION> tag at
   the end. Also the cogl-vector section was added in the middle of the
   cogl-buffer one. Let's shiffle it out and add that </SECTION> tag.
2010-06-07 14:32:32 +01:00
Damien Lespiau
7fa011b037 build: Don't distribute clutter-version.h and cogl-defines.h
As with a351ff2af earlier, distributing headers generated at configure
time conflicts with out of tree builds as the distributed headers will
be included first instead of including the generated ones.
2010-06-04 17:43:57 +01:00
Robert Bragg
7dbcb82622 cogl-object-private.h: white space cleanup
simply replaces tabs with spaces
2010-06-04 14:44:15 +01:00
Robert Bragg
d02d40bf61 CoglObject: Adds a COGL_OBJECT macro for casting
This adds a convenience COGL_OBJECT macro for casting CoglObject
subclasses to (CoglObject *) pointers.
2010-06-04 14:44:15 +01:00
Robert Bragg
5af3ead3a2 CoglObject: Adds cogl_object_{get,set}_user_data
This provides a mechanism for associating private data with any
CoglObject. We expect Clutter will use this to associate weak materials
with normal materials.
2010-06-04 14:44:15 +01:00
Robert Bragg
ef08c6369a cogl-path: Adds missing cogl_is_path prototype
This adds a missing cogl_is_path prototype to the public cogl-path.h
header.
2010-06-04 14:44:15 +01:00
Robert Bragg
368dc48372 cogl-path: Renames cogl_path_get/set cogl_get/set_path
These aren't path methods so aren't consistent with the
cogl_object_method naming style we are aiming for.
2010-06-04 14:44:15 +01:00
Damien Lespiau
63eeaa86ea build: Distribute cogl-object.h
cogl-object.h is a new public header. Include it in the right list of
source files to have it distributed.
2010-06-03 11:41:03 +01:00
Emmanuele Bassi
cd7c1dca76 docs: Miscellaneous fixes to silence gtk-doc 2010-06-01 14:15:03 +01:00
Robert Bragg
595271ac87 clip_stack: Remove use of CoglHandle in the CoglClipStack API
This replaces the use of CoglHandle with strongly type CoglClipStack *
pointers instead. The only function not converted for now is
cogl_is_clip_stack which will be done in a later commit.
2010-06-01 12:20:59 +01:00
Robert Bragg
653d1a3254 bitmap: Remove use of CoglHandle in the CoglBitmap API
This replaces the use of CoglHandle with strongly type CoglBitmap *
pointers instead. The only function not converted for now is
cogl_is_bitmap which will be done in a later commit.
2010-06-01 12:20:59 +01:00
Robert Bragg
817c1cddcc path: Remove use of CoglHandle in the CoglPath API
This replaces the use of CoglHandle with strongly type CoglPath *
pointers instead. The only function not converted for now is
cogl_is_path which will be done in a later commit.
2010-06-01 12:20:58 +01:00
Robert Bragg
5f8fadeb1c cogl: avoid ‘_cogl_xyz_handle_new’ defined but not used warnings
This patch makes it so that only the backwards compatibility
COGL_HANDLE_DEFINE macro defines a _cogl_xyz_handle_new function. The
new COGL_OBJECT_DEFINE macro only defines a _cogl_xyz_object_new
function.
2010-06-01 12:20:58 +01:00
Damien Lespiau
db7cc5b1b4 analysis: Fix some CoglFixed API
2 functions where not correctly defined/documented, use the actual
symbol name instead.
2010-06-01 12:08:18 +01:00
Damien Lespiau
45a947d226 analysis: Use static functions when possible
It's best to ensure the functions only used in one compilation unit are
marked as static so GCC can inline them if it feels like it.
2010-06-01 12:08:18 +01:00
Damien Lespiau
338a37d468 analysis: Make all function declarations be protoypes
It's valid C to declare a function omitting it prototype, but it seems
to be a good practise to always declare a function with its
corresponding prototype.
2010-06-01 12:08:18 +01:00
Damien Lespiau
58b0028b52 analysis: FALSE/0 used in pointer context
While this is totally fine (0 in the pointer context will be converted
in the right internal NULL representation, which could be a value with
some bits to 1), I believe it's clearer to use NULL in the pointer
context.

It seems that, in most case, it's more an overlook than a deliberate
choice to use FALSE/0 as NULL, eg. copying a _COGL_GET_CONTEXT (ctx, 0)
or a g_return_val_if_fail (cond, 0) from a function returning a
gboolean.
2010-06-01 12:08:18 +01:00
Robert Bragg
3aaef72e50 buffer: Remove use of CoglHandle in the CoglBuffer API
This replaces the use of CoglHandle with strongly type CoglBuffer *
pointers instead. The only function not converted for now is
cogl_is_buffer which will be done in a later commit.
2010-05-28 18:42:33 +01:00
Robert Bragg
7bcea1c64e Start eradicating the CoglHandle typedef
CoglHandle is a common source of complaints and confusion because people
expect a "handle" to be some form of integer type with some indirection
to lookup the corresponding objects as opposed to a direct pointer.

This patch starts by renaming CoglHandle to CoglObject * and creating
corresponding cogl_object_ APIs to replace the cogl_handle ones.

The next step though is to remove all use of CoglHandle in the Cogl APIs
and replace with strongly typed pointer types such as CoglMaterial * or
CoglTexture * etc also all occurrences of COGL_INVALID_HANDLE can just
use NULL instead.

After this we will consider switching to GTypeInstance internally so we
can have inheritance for our types and hopefully improve how we handle
bindings.

Note all these changes will be done in a way that maintains the API and
ABI.
2010-05-28 18:39:22 +01:00
Neil Roberts
0f32fb4acd cogl-texture-2d: Regenerate the mipmaps when the texture is modified
When the texture is modified by cogl_texture_set_region, it wasn't
setting mipmaps_dirty so it would never update the mipmaps.
2010-05-28 18:13:53 +01:00
Robert Bragg
4c1a12d824 [material] fix _set_alpha_test_function: missing a glEnable(GL_ALPHA_TEST)
Since the default alpha test function of GL_ALWAYS is equivalent to
GL_ALPHA_TEST being disabled we don't need to worry about Enabling/Disabling
it when flushing material state, instead it's enough to leave it always
enabled.  We will assume that any driver worth its salt wont incur any
additional cost for glEnable (GL_ALPHA_TEST) + GL_ALWAYS vs
glDisable (GL_ALPHA_TEST).

This patch simply calls glEnable (GL_ALPHA_TEST) in cogl_create_context
2010-05-28 15:35:23 +01:00
Neil Roberts
5a1cffd0bb cogl-vertex-buffer: Don't disable any texture coord arrays
The enabled state of the texture coord arrays is cached in the cogl
context so there is no need to disable them when we've finished
rendering a vbo.
2010-05-27 14:54:14 +01:00