mutter/cogl/cogl-primitives.c

1238 lines
44 KiB
C
Raw Normal View History

/*
* Cogl
*
* An object oriented GL/GLES Abstraction/Utility Layer
*
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
* Copyright (C) 2007,2008,2009,2010 Intel Corporation.
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library. If not, see <http://www.gnu.org/licenses/>.
*
*
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "cogl.h"
#include "cogl-internal.h"
#include "cogl-context.h"
#include "cogl-journal-private.h"
#include "cogl-texture-private.h"
#include "cogl-material-private.h"
#include "cogl-vertex-buffer-private.h"
#include "cogl-framebuffer-private.h"
#include <string.h>
#include <math.h>
#define _COGL_MAX_BEZ_RECURSE_DEPTH 16
#ifdef HAVE_COGL_GL
#define glClientActiveTexture ctx->drv.pf_glClientActiveTexture
#endif
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
typedef struct _TextureSlicedQuadState
{
CoglHandle material;
float tex_virtual_origin_x;
float tex_virtual_origin_y;
float quad_origin_x;
float quad_origin_y;
float v_to_q_scale_x;
float v_to_q_scale_y;
float quad_len_x;
float quad_len_y;
gboolean flipped_x;
gboolean flipped_y;
CoglMaterialWrapModeOverrides *wrap_mode_overrides;
} TextureSlicedQuadState;
typedef struct _TextureSlicedPolygonState
{
const CoglTextureVertex *vertices;
int n_vertices;
int stride;
} TextureSlicedPolygonState;
static void
log_quad_sub_textures_cb (CoglHandle texture_handle,
GLuint gl_handle,
GLenum gl_target,
const float *subtexture_coords,
const float *virtual_coords,
void *user_data)
{
TextureSlicedQuadState *state = user_data;
float quad_coords[4];
#define TEX_VIRTUAL_TO_QUAD(V, Q, AXIS) \
do { \
Q = V - state->tex_virtual_origin_##AXIS; \
Q *= state->v_to_q_scale_##AXIS; \
if (state->flipped_##AXIS) \
Q = state->quad_len_##AXIS - Q; \
Q += state->quad_origin_##AXIS; \
} while (0);
TEX_VIRTUAL_TO_QUAD (virtual_coords[0], quad_coords[0], x);
TEX_VIRTUAL_TO_QUAD (virtual_coords[1], quad_coords[1], y);
TEX_VIRTUAL_TO_QUAD (virtual_coords[2], quad_coords[2], x);
TEX_VIRTUAL_TO_QUAD (virtual_coords[3], quad_coords[3], y);
#undef TEX_VIRTUAL_TO_QUAD
COGL_NOTE (DRAW,
"~~~~~ slice\n"
"qx1: %f\t"
"qy1: %f\n"
"qx2: %f\t"
"qy2: %f\n"
"tx1: %f\t"
"ty1: %f\n"
"tx2: %f\t"
"ty2: %f\n",
quad_coords[0], quad_coords[1],
quad_coords[2], quad_coords[3],
subtexture_coords[0], subtexture_coords[1],
subtexture_coords[2], subtexture_coords[3]);
/* FIXME: when the wrap mode becomes part of the material we need to
* be able to override the wrap mode when logging a quad. */
_cogl_journal_log_quad (quad_coords,
state->material,
1, /* one layer */
0, /* don't need to use fallbacks */
gl_handle, /* replace the layer0 texture */
state->wrap_mode_overrides, /* use GL_CLAMP_TO_EDGE */
subtexture_coords,
4);
}
/* This path doesn't currently support multitexturing but is used for
* CoglTextures that don't support repeating using the GPU so we need to
* manually emit extra geometry to fake the repeating. This includes:
*
* - CoglTexture2DSliced: when made of > 1 slice or if the users given
* texture coordinates require repeating,
* - CoglTexture2DAtlas: if the users given texture coordinates require
* repeating,
* - CoglTextureRectangle: if the users given texture coordinates require
* repeating,
* - CoglTexturePixmap: if the users given texture coordinates require
* repeating
*/
/* TODO: support multitexturing */
static void
_cogl_texture_quad_multiple_primitives (CoglHandle tex_handle,
CoglHandle material,
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
gboolean clamp_s,
gboolean clamp_t,
const float *position,
float tx_1,
float ty_1,
float tx_2,
float ty_2)
{
TextureSlicedQuadState state;
CoglMaterialWrapModeOverrides wrap_mode_overrides;
gboolean tex_virtual_flipped_x;
gboolean tex_virtual_flipped_y;
gboolean quad_flipped_x;
gboolean quad_flipped_y;
CoglHandle first_layer;
_COGL_GET_CONTEXT (ctx, NO_RETVAL);
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
/* If the wrap mode is clamp to edge then we'll recursively draw the
stretched part and replace the coordinates */
if (clamp_s && tx_1 != tx_2)
{
float *replacement_position = g_newa (float, 4);
float old_tx_1 = tx_1, old_tx_2 = tx_2;
memcpy (replacement_position, position, sizeof (float) * 4);
tx_1 = CLAMP (tx_1, 0.0f, 1.0f);
tx_2 = CLAMP (tx_2, 0.0f, 1.0f);
if (old_tx_1 != tx_1)
{
/* Draw the left part of the quad as a stretched copy of tx_1 */
float tmp_position[] =
{ position[0], position[1],
(position[0] +
(position[2] - position[0]) *
(tx_1 - old_tx_1) / (old_tx_2 - old_tx_1)),
position[3] };
_cogl_texture_quad_multiple_primitives (tex_handle, material,
FALSE, clamp_t,
tmp_position,
tx_1, ty_1, tx_1, ty_2);
replacement_position[0] = tmp_position[2];
}
if (old_tx_2 != tx_2)
{
/* Draw the right part of the quad as a stretched copy of tx_2 */
float tmp_position[] =
{ (position[0] +
(position[2] - position[0]) *
(tx_2 - old_tx_1) / (old_tx_2 - old_tx_1)),
position[1], position[2], position[3] };
_cogl_texture_quad_multiple_primitives (tex_handle, material,
FALSE, clamp_t,
tmp_position,
tx_2, ty_1, tx_2, ty_2);
replacement_position[2] = tmp_position[0];
}
/* If there's no main part left then we don't need to continue */
if (tx_1 == tx_2)
return;
position = replacement_position;
}
if (clamp_t && ty_1 != ty_2)
{
float *replacement_position = g_newa (float, 4);
float old_ty_1 = ty_1, old_ty_2 = ty_2;
memcpy (replacement_position, position, sizeof (float) * 4);
ty_1 = CLAMP (ty_1, 0.0f, 1.0f);
ty_2 = CLAMP (ty_2, 0.0f, 1.0f);
if (old_ty_1 != ty_1)
{
/* Draw the top part of the quad as a stretched copy of ty_1 */
float tmp_position[] =
{ position[0], position[1], position[2],
(position[1] +
(position[3] - position[1]) *
(ty_1 - old_ty_1) / (old_ty_2 - old_ty_1)) };
_cogl_texture_quad_multiple_primitives (tex_handle, material,
clamp_s, FALSE,
tmp_position,
tx_1, ty_1, tx_2, ty_1);
replacement_position[1] = tmp_position[3];
}
if (old_ty_2 != ty_2)
{
/* Draw the bottom part of the quad as a stretched copy of ty_2 */
float tmp_position[] =
{ position[0],
(position[1] +
(position[3] - position[1]) *
(ty_2 - old_ty_1) / (old_ty_2 - old_ty_1)),
position[2], position[3] };
_cogl_texture_quad_multiple_primitives (tex_handle, material,
clamp_s, FALSE,
tmp_position,
tx_1, ty_2, tx_2, ty_2);
replacement_position[3] = tmp_position[1];
}
/* If there's no main part left then we don't need to continue */
if (ty_1 == ty_2)
return;
position = replacement_position;
}
state.wrap_mode_overrides = NULL;
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
memset (&state.wrap_mode_overrides, 0, sizeof (state.wrap_mode_overrides));
/* We can't use hardware repeat so we need to set clamp to edge
otherwise it might pull in edge pixels from the other side. By
default WRAP_MODE_AUTOMATIC becomes CLAMP_TO_EDGE so we only need
to override if the wrap mode is repeat */
first_layer = cogl_material_get_layers (material)->data;
if (cogl_material_layer_get_wrap_mode_s (first_layer) ==
COGL_MATERIAL_WRAP_MODE_REPEAT)
{
state.wrap_mode_overrides = &wrap_mode_overrides;
wrap_mode_overrides.values[0].s =
COGL_MATERIAL_WRAP_MODE_OVERRIDE_CLAMP_TO_EDGE;
}
if (cogl_material_layer_get_wrap_mode_t (first_layer) ==
COGL_MATERIAL_WRAP_MODE_REPEAT)
{
state.wrap_mode_overrides = &wrap_mode_overrides;
wrap_mode_overrides.values[0].t =
COGL_MATERIAL_WRAP_MODE_OVERRIDE_CLAMP_TO_EDGE;
}
state.material = material;
/* Get together the data we need to transform the virtual texture
* coordinates of each slice into quad coordinates...
*
* NB: We need to consider that the quad coordinates and the texture
* coordinates may be inverted along the x or y axis, and must preserve the
* inversions when we emit the final geometry.
*/
#define X0 0
#define Y0 1
#define X1 2
#define Y1 3
tex_virtual_flipped_x = (tx_1 > tx_2) ? TRUE : FALSE;
tex_virtual_flipped_y = (ty_1 > ty_2) ? TRUE : FALSE;
state.tex_virtual_origin_x = tex_virtual_flipped_x ? tx_2 : tx_1;
state.tex_virtual_origin_y = tex_virtual_flipped_y ? ty_2 : ty_1;
quad_flipped_x = (position[X0] > position[X1]) ? TRUE : FALSE;
quad_flipped_y = (position[Y0] > position[Y1]) ? TRUE : FALSE;
state.quad_origin_x = quad_flipped_x ? position[X1] : position[X0];
state.quad_origin_y = quad_flipped_y ? position[Y1] : position[Y0];
/* flatten the two forms of coordinate inversion into one... */
state.flipped_x = tex_virtual_flipped_x ^ quad_flipped_x;
state.flipped_y = tex_virtual_flipped_y ^ quad_flipped_y;
/* We use the _len_AXIS naming here instead of _width and _height because
* log_quad_slice_cb uses a macro with symbol concatenation to handle both
* axis, so this is more convenient... */
state.quad_len_x = fabs (position[X1] - position[X0]);
state.quad_len_y = fabs (position[Y1] - position[Y0]);
#undef X0
#undef Y0
#undef X1
#undef Y1
state.v_to_q_scale_x = fabs (state.quad_len_x / (tx_2 - tx_1));
state.v_to_q_scale_y = fabs (state.quad_len_y / (ty_2 - ty_1));
_cogl_texture_foreach_sub_texture_in_region (tex_handle,
tx_1, ty_1, tx_2, ty_2,
log_quad_sub_textures_cb,
&state);
}
/* This path supports multitexturing but only when each of the layers is
* handled with a single GL texture. Also if repeating is necessary then
* _cogl_texture_can_hardware_repeat() must return TRUE.
* This includes layers made from:
*
* - CoglTexture2DSliced: if only comprised of a single slice with optional
* waste, assuming the users given texture coordinates don't require
* repeating.
* - CoglTexture{1D,2D,3D}: always.
* - CoglTexture2DAtlas: assuming the users given texture coordinates don't
* require repeating.
* - CoglTextureRectangle: assuming the users given texture coordinates don't
* require repeating.
* - CoglTexturePixmap: assuming the users given texture coordinates don't
* require repeating.
*/
static gboolean
_cogl_multitexture_quad_single_primitive (const float *position,
CoglHandle material,
guint32 fallback_layers,
const float *user_tex_coords,
int user_tex_coords_len)
{
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
int n_layers = cogl_material_get_n_layers (material);
float *final_tex_coords = alloca (sizeof (float) * 4 * n_layers);
const GList *layers;
GList *tmp;
int i;
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
CoglMaterialWrapModeOverrides wrap_mode_overrides;
/* This will be set to point to wrap_mode_overrides when an override
is needed */
CoglMaterialWrapModeOverrides *wrap_mode_overrides_p = NULL;
_COGL_GET_CONTEXT (ctx, FALSE);
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
memset (&wrap_mode_overrides, 0, sizeof (wrap_mode_overrides));
/*
* Validate the texture coordinates for this rectangle.
*/
layers = cogl_material_get_layers (material);
for (tmp = (GList *)layers, i = 0; tmp != NULL; tmp = tmp->next, i++)
{
CoglHandle layer = (CoglHandle)tmp->data;
CoglHandle tex_handle;
const float *in_tex_coords;
float *out_tex_coords;
float default_tex_coords[4] = {0.0, 0.0, 1.0, 1.0};
CoglTransformResult transform_result;
tex_handle = cogl_material_layer_get_texture (layer);
/* COGL_INVALID_HANDLE textures are handled by
* _cogl_material_flush_gl_state */
if (tex_handle == COGL_INVALID_HANDLE)
continue;
/* If the user didn't supply texture coordinates for this layer
then use the default coords */
if (i >= user_tex_coords_len / 4)
in_tex_coords = default_tex_coords;
else
in_tex_coords = &user_tex_coords[i * 4];
out_tex_coords = &final_tex_coords[i * 4];
memcpy (out_tex_coords, in_tex_coords, sizeof (GLfloat) * 4);
/* Convert the texture coordinates to GL.
*/
transform_result =
_cogl_texture_transform_quad_coords_to_gl (tex_handle,
out_tex_coords);
/* If the texture has waste or we are using GL_TEXTURE_RECT we
* can't handle texture repeating so we can't use the layer if
* repeating is required.
*
* NB: We already know that no texture matrix is being used if the
* texture doesn't support hardware repeat.
*/
if (transform_result == COGL_TRANSFORM_SOFTWARE_REPEAT)
{
if (i == 0)
{
if (n_layers > 1)
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
g_warning ("Skipping layers 1..n of your material since "
"the first layer doesn't support hardware "
"repeat (e.g. because of waste or use of "
"GL_TEXTURE_RECTANGLE_ARB) and you supplied "
"texture coordinates outside the range [0,1]."
"Falling back to software repeat assuming "
"layer 0 is the most important one keep");
warning_seen = TRUE;
}
return FALSE;
}
else
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
g_warning ("Skipping layer %d of your material "
"since you have supplied texture coords "
"outside the range [0,1] but the texture "
"doesn't support hardware repeat (e.g. "
"because of waste or use of "
"GL_TEXTURE_RECTANGLE_ARB). This isn't "
"supported with multi-texturing.", i);
warning_seen = TRUE;
/* NB: marking for fallback will replace the layer with
* a default transparent texture */
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
fallback_layers |= (1 << i);
}
}
/* By default WRAP_MODE_AUTOMATIC becomes to CLAMP_TO_EDGE. If
the texture coordinates need repeating then we'll override
this to GL_REPEAT. Otherwise we'll leave it at CLAMP_TO_EDGE
so that it won't blend in pixels from the opposite side when
the full texture is drawn with GL_LINEAR filter mode */
if (transform_result == COGL_TRANSFORM_HARDWARE_REPEAT)
{
if (cogl_material_layer_get_wrap_mode_s (layer) ==
COGL_MATERIAL_WRAP_MODE_AUTOMATIC)
{
wrap_mode_overrides.values[i].s
= COGL_MATERIAL_WRAP_MODE_OVERRIDE_REPEAT;
wrap_mode_overrides_p = &wrap_mode_overrides;
}
if (cogl_material_layer_get_wrap_mode_t (layer) ==
COGL_MATERIAL_WRAP_MODE_AUTOMATIC)
{
wrap_mode_overrides.values[i].t
= COGL_MATERIAL_WRAP_MODE_OVERRIDE_REPEAT;
wrap_mode_overrides_p = &wrap_mode_overrides;
}
}
}
_cogl_journal_log_quad (position,
material,
n_layers,
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
fallback_layers,
0, /* don't replace the layer0 texture */
wrap_mode_overrides_p,
final_tex_coords,
n_layers * 4);
return TRUE;
}
struct _CoglMutiTexturedRect
{
const float *position; /* x0,y0,x1,y1 */
const float *tex_coords; /* (tx0,ty0,tx1,ty1)(tx0,ty0,tx1,ty1)(... */
int tex_coords_len; /* number of floats in tex_coords? */
};
static void
_cogl_rectangles_with_multitexture_coords (
struct _CoglMutiTexturedRect *rects,
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
int n_rects)
{
CoglHandle material;
const GList *layers;
int n_layers;
const GList *tmp;
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
guint32 fallback_layers = 0;
gboolean all_use_sliced_quad_fallback = FALSE;
int i;
_COGL_GET_CONTEXT (ctx, NO_RETVAL);
material = ctx->source_material;
layers = cogl_material_get_layers (material);
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
n_layers = cogl_material_get_n_layers (material);
/*
* Validate all the layers of the current source material...
*/
for (tmp = layers, i = 0; tmp != NULL; tmp = tmp->next, i++)
{
CoglHandle layer = tmp->data;
CoglHandle tex_handle;
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
unsigned long flags;
if (cogl_material_layer_get_type (layer)
!= COGL_MATERIAL_LAYER_TYPE_TEXTURE)
continue;
/* We need to ensure the mipmaps are ready before deciding
* anything else about the texture because the texture storage
* could completely change if it needs to be migrated out of the
* atlas and will affect how we validate the layer.
*
* FIXME: this needs to be generalized. There could be any
* number of things that might require a shuffling of the
* underlying texture storage. We could add two mechanisms to
* generalize this a bit...
*
* 1) add a _cogl_material_layer_update_storage() function that
* would for instance consider if mipmapping is necessary and
* potentially migrate the texture from an atlas.
*
* 2) allow setting of transient primitive-flags on a material
* that may affect the outcome of _update_storage(). One flag
* could indicate that we expect to sample beyond the bounds of
* the texture border.
*
* flags = COGL_MATERIAL_PRIMITIVE_FLAG_VALID_BORDERS;
* _cogl_material_layer_assert_primitive_flags (layer, flags)
* _cogl_material_layer_update_storage (layer)
* enqueue primitive in journal
*
* when the primitive is dequeued and drawn we should:
* _cogl_material_flush_gl_state (material)
* draw primitive
* _cogl_material_unassert_primitive_flags (layer, flags);
*
* _cogl_material_layer_update_storage should take into
* consideration all the asserted primitive requirements. (E.g.
* there could be multiple primitives in the journal - or in a
* renderlist in the future - that need mipmaps or that need
* valid contents beyond their borders (for cogl_polygon)
* meaning they can't work with textures in an atas, so
* _cogl_material_layer_update_storage would pass on these
* requirements to the texture atlas backend which would make
* sure the referenced texture is migrated out of the atlas and
* mipmaps are generated.)
*/
_cogl_material_layer_ensure_mipmaps (layer);
tex_handle = cogl_material_layer_get_texture (layer);
/* COGL_INVALID_HANDLE textures are handled by
* _cogl_material_flush_gl_state */
if (tex_handle == COGL_INVALID_HANDLE)
continue;
/* XXX:
* For now, if the first layer is sliced then all other layers are
* ignored since we currently don't support multi-texturing with
* sliced textures. If the first layer is not sliced then any other
* layers found to be sliced will be skipped. (with a warning)
*
* TODO: Add support for multi-texturing rectangles with sliced
* textures if no texture matrices are in use.
*/
if (cogl_texture_is_sliced (tex_handle))
{
if (i == 0)
{
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
fallback_layers = ~1; /* fallback all except the first layer */
all_use_sliced_quad_fallback = TRUE;
if (tmp->next)
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
g_warning ("Skipping layers 1..n of your material since "
"the first layer is sliced. We don't currently "
"support any multi-texturing with sliced "
"textures but assume layer 0 is the most "
"important to keep");
warning_seen = TRUE;
}
break;
}
else
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
g_warning ("Skipping layer %d of your material consisting of "
"a sliced texture (unsuported for multi texturing)",
i);
warning_seen = TRUE;
/* NB: marking for fallback will replace the layer with
* a default transparent texture */
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
fallback_layers |= (1 << i);
continue;
}
}
/* If the texture can't be repeated with the GPU (e.g. because it has
* waste or if using GL_TEXTURE_RECTANGLE_ARB) then we don't support
* multi texturing since we don't know if the result will end up trying
* to texture from the waste area. */
flags = _cogl_material_layer_get_flags (layer);
if (flags & COGL_MATERIAL_LAYER_FLAG_HAS_USER_MATRIX
&& !_cogl_texture_can_hardware_repeat (tex_handle))
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
g_warning ("Skipping layer %d of your material since a custom "
"texture matrix was given for a texture that can't be "
"repeated using the GPU and the result may try to "
"sample beyond the bounds of the texture ",
i);
warning_seen = TRUE;
/* NB: marking for fallback will replace the layer with
* a default transparent texture */
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
fallback_layers |= (1 << i);
continue;
}
}
/*
* Emit geometry for each of the rectangles...
*/
for (i = 0; i < n_rects; i++)
{
CoglHandle first_layer, tex_handle;
const float default_tex_coords[4] = {0.0, 0.0, 1.0, 1.0};
const float *tex_coords;
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
gboolean clamp_s, clamp_t;
if (!all_use_sliced_quad_fallback)
{
gboolean success =
_cogl_multitexture_quad_single_primitive (rects[i].position,
material,
fallback_layers,
rects[i].tex_coords,
rects[i].tex_coords_len);
/* NB: If _cogl_multitexture_quad_single_primitive fails then it
* means the user tried to use texture repeat with a texture that
* can't be repeated by the GPU (e.g. due to waste or use of
* GL_TEXTURE_RECTANGLE_ARB) */
if (success)
continue;
}
/* If multitexturing failed or we are drawing with a sliced texture
* then we only support a single layer so we pluck out the texture
* from the first material layer... */
first_layer = layers->data;
tex_handle = cogl_material_layer_get_texture (first_layer);
if (rects[i].tex_coords)
tex_coords = rects[i].tex_coords;
else
tex_coords = default_tex_coords;
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
clamp_s = (cogl_material_layer_get_wrap_mode_s (first_layer) ==
COGL_MATERIAL_WRAP_MODE_CLAMP_TO_EDGE);
clamp_t = (cogl_material_layer_get_wrap_mode_t (first_layer) ==
COGL_MATERIAL_WRAP_MODE_CLAMP_TO_EDGE);
COGL_NOTE (DRAW, "Drawing Tex Quad (Multi-Prim Mode)");
_cogl_texture_quad_multiple_primitives (tex_handle,
material,
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
clamp_s, clamp_t,
rects[i].position,
tex_coords[0],
tex_coords[1],
tex_coords[2],
tex_coords[3]);
}
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
#if 0
/* XXX: The current journal doesn't handle changes to the model view matrix
* so for now we force a flush at the end of every primitive. */
_cogl_journal_flush ();
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
#endif
}
void
cogl_rectangles (const float *verts,
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
unsigned int n_rects)
{
struct _CoglMutiTexturedRect *rects;
int i;
/* XXX: All the cogl_rectangle* APIs normalize their input into an array of
* _CoglMutiTexturedRect rectangles and pass these on to our work horse;
* _cogl_rectangles_with_multitexture_coords.
*/
rects = g_alloca (n_rects * sizeof (struct _CoglMutiTexturedRect));
for (i = 0; i < n_rects; i++)
{
rects[i].position = &verts[i * 4];
rects[i].tex_coords = NULL;
rects[i].tex_coords_len = 0;
}
_cogl_rectangles_with_multitexture_coords (rects, n_rects);
}
void
cogl_rectangles_with_texture_coords (const float *verts,
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
unsigned int n_rects)
{
struct _CoglMutiTexturedRect *rects;
int i;
/* XXX: All the cogl_rectangle* APIs normalize their input into an array of
* _CoglMutiTexturedRect rectangles and pass these on to our work horse;
* _cogl_rectangles_with_multitexture_coords.
*/
rects = g_alloca (n_rects * sizeof (struct _CoglMutiTexturedRect));
for (i = 0; i < n_rects; i++)
{
rects[i].position = &verts[i * 8];
rects[i].tex_coords = &verts[i * 8 + 4];
rects[i].tex_coords_len = 4;
}
_cogl_rectangles_with_multitexture_coords (rects, n_rects);
}
void
cogl_rectangle_with_texture_coords (float x_1,
float y_1,
float x_2,
float y_2,
float tx_1,
float ty_1,
float tx_2,
float ty_2)
{
const float position[4] = {x_1, y_1, x_2, y_2};
const float tex_coords[4] = {tx_1, ty_1, tx_2, ty_2};
struct _CoglMutiTexturedRect rect;
/* XXX: All the cogl_rectangle* APIs normalize their input into an array of
* _CoglMutiTexturedRect rectangles and pass these on to our work horse;
* _cogl_rectangles_with_multitexture_coords.
*/
rect.position = position;
rect.tex_coords = tex_coords;
rect.tex_coords_len = 4;
_cogl_rectangles_with_multitexture_coords (&rect, 1);
}
void
cogl_rectangle_with_multitexture_coords (float x_1,
float y_1,
float x_2,
float y_2,
const float *user_tex_coords,
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
int user_tex_coords_len)
{
const float position[4] = {x_1, y_1, x_2, y_2};
struct _CoglMutiTexturedRect rect;
/* XXX: All the cogl_rectangle* APIs normalize their input into an array of
* _CoglMutiTexturedRect rectangles and pass these on to our work horse;
* _cogl_rectangles_with_multitexture_coords.
*/
rect.position = position;
rect.tex_coords = user_tex_coords;
rect.tex_coords_len = user_tex_coords_len;
_cogl_rectangles_with_multitexture_coords (&rect, 1);
}
void
cogl_rectangle (float x_1,
float y_1,
float x_2,
float y_2)
{
const float position[4] = {x_1, y_1, x_2, y_2};
struct _CoglMutiTexturedRect rect;
/* XXX: All the cogl_rectangle* APIs normalize their input into an array of
* _CoglMutiTexturedRect rectangles and pass these on to our work horse;
* _cogl_rectangles_with_multitexture_coords.
*/
rect.position = position;
rect.tex_coords = NULL;
rect.tex_coords_len = 0;
_cogl_rectangles_with_multitexture_coords (&rect, 1);
}
void
draw_polygon_sub_texture_cb (CoglHandle tex_handle,
GLuint gl_handle,
GLenum gl_target,
const float *subtexture_coords,
const float *virtual_coords,
void *user_data)
{
TextureSlicedPolygonState *state = user_data;
GLfloat *v;
int i;
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
CoglMaterialFlushOptions options;
float slice_origin_x;
float slice_origin_y;
float virtual_origin_x;
float virtual_origin_y;
float v_to_s_scale_x;
float v_to_s_scale_y;
_COGL_GET_CONTEXT (ctx, NO_RETVAL);
slice_origin_x = subtexture_coords[0];
slice_origin_y = subtexture_coords[1];
virtual_origin_x = virtual_coords[0];
virtual_origin_y = virtual_coords[1];
v_to_s_scale_x = ((virtual_coords[2] - virtual_coords[0]) /
(subtexture_coords[2] - subtexture_coords[0]));
v_to_s_scale_y = ((virtual_coords[3] - virtual_coords[1]) /
(subtexture_coords[3] - subtexture_coords[1]));
/* Convert the vertices into an array of GLfloats ready to pass to
* OpenGL */
v = (GLfloat *)ctx->logged_vertices->data;
for (i = 0; i < state->n_vertices; i++)
{
/* NB: layout = [X,Y,Z,TX,TY,R,G,B,A,...] */
GLfloat *t = v + 3;
t[0] = ((state->vertices[i].tx - virtual_origin_x) * v_to_s_scale_x
+ slice_origin_x);
t[1] = ((state->vertices[i].ty - virtual_origin_y) * v_to_s_scale_y
+ slice_origin_y);
v += state->stride;
}
options.flags =
COGL_MATERIAL_FLUSH_DISABLE_MASK |
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
COGL_MATERIAL_FLUSH_LAYER0_OVERRIDE |
COGL_MATERIAL_FLUSH_WRAP_MODE_OVERRIDES;
/* disable all except the first layer */
options.disable_layers = (guint32)~1;
options.layer0_override_texture = gl_handle;
cogl-material: Add support for setting the wrap mode for a layer Previously, Cogl's texture coordinate system was effectively always GL_REPEAT so that if an application specifies coordinates outside the range 0→1 it would get repeated copies of the texture. It would however change the mode to GL_CLAMP_TO_EDGE if all of the coordinates are in the range 0→1 so that in the common case that the whole texture is being drawn with linear filtering it will not blend in edge pixels from the opposite sides. This patch adds the option for applications to change the wrap mode per layer. There are now three wrap modes: 'repeat', 'clamp-to-edge' and 'automatic'. The automatic map mode is the default and it implements the previous behaviour. The wrap mode can be changed for the s and t coordinates independently. I've tried to make the internals support setting the r coordinate but as we don't support 3D textures yet I haven't exposed any public API for it. The texture backends still have a set_wrap_mode virtual but this value is intended to be transitory and it will be changed whenever the material is flushed (although the backends are expected to cache it so that it won't use too many GL calls). In my understanding this value was always meant to be transitory and all primitives were meant to set the value before drawing. However there were comments suggesting that this is not the expected behaviour. In particular the vertex buffer drawing code never set a wrap mode so it would end up with whatever the texture was previously used for. These issues are now fixed because the material will always set the wrap modes. There is code to manually implement clamp-to-edge for textures that can't be hardware repeated. However this doesn't fully work because it relies on being able to draw the stretched parts using quads with the same values for tx1 and tx2. The texture iteration code doesn't support this so it breaks. This is a separate bug and it isn't trivially solved. When flushing a material there are now extra options to set wrap mode overrides. The overrides are an array of values for each layer that specifies an override for the s, t or r coordinates. The primitives use this to implement the automatic wrap mode. cogl_polygon also uses it to set GL_CLAMP_TO_BORDER mode for its trick to render sliced textures. Although this code has been added it looks like the sliced trick has been broken for a while and I haven't attempted to fix it here. I've added a constant to represent the maximum number of layers that a material supports so that I can size the overrides array. I've set it to 32 because as far as I can tell we have that limit imposed anyway because the other flush options use a guint32 to store a flag about each layer. The overrides array ends up adding 32 bytes to each flush options struct which may be a concern. http://bugzilla.openedhand.com/show_bug.cgi?id=2063
2010-04-01 06:31:33 -04:00
/* Override the wrapping mode on all of the slices to use a
transparent border so that we can draw the full polygon for
each slice. Coordinates outside the texture will be transparent
so only the part of the polygon that intersects the slice will
be visible. This is a fairly hacky fallback and it relies on
the blending function working correctly */
memset (&options.wrap_mode_overrides, 0,
sizeof (options.wrap_mode_overrides));
options.wrap_mode_overrides.values[0].s =
COGL_MATERIAL_WRAP_MODE_OVERRIDE_CLAMP_TO_BORDER;
options.wrap_mode_overrides.values[0].t =
COGL_MATERIAL_WRAP_MODE_OVERRIDE_CLAMP_TO_BORDER;
_cogl_material_flush_gl_state (ctx->source_material, &options);
GE (glDrawArrays (GL_TRIANGLE_FAN, 0, state->n_vertices));
}
/* handles 2d-sliced textures with > 1 slice */
static void
_cogl_texture_polygon_multiple_primitives (const CoglTextureVertex *vertices,
unsigned int n_vertices,
unsigned int stride,
gboolean use_color)
{
const GList *layers;
CoglHandle layer0;
CoglHandle tex_handle;
GLfloat *v;
int i;
TextureSlicedPolygonState state;
_COGL_GET_CONTEXT (ctx, NO_RETVAL);
/* We can assume in this case that we have at least one layer in the
* material that corresponds to a sliced cogl texture */
layers = cogl_material_get_layers (ctx->source_material);
layer0 = (CoglHandle)layers->data;
tex_handle = cogl_material_layer_get_texture (layer0);
v = (GLfloat *)ctx->logged_vertices->data;
for (i = 0; i < n_vertices; i++)
{
guint8 *c;
v[0] = vertices[i].x;
v[1] = vertices[i].y;
v[2] = vertices[i].z;
if (use_color)
{
/* NB: [X,Y,Z,TX,TY,R,G,B,A,...] */
c = (guint8 *) (v + 5);
c[0] = cogl_color_get_red_byte (&vertices[i].color);
c[1] = cogl_color_get_green_byte (&vertices[i].color);
c[2] = cogl_color_get_blue_byte (&vertices[i].color);
c[3] = cogl_color_get_alpha_byte (&vertices[i].color);
}
v += stride;
}
state.stride = stride;
state.vertices = vertices;
state.n_vertices = n_vertices;
_cogl_texture_foreach_sub_texture_in_region (tex_handle,
0, 0, 1, 1,
draw_polygon_sub_texture_cb,
&state);
}
static void
_cogl_multitexture_polygon_single_primitive (const CoglTextureVertex *vertices,
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
unsigned int n_vertices,
unsigned int n_layers,
unsigned int stride,
gboolean use_color,
guint32 fallback_layers,
CoglMaterialWrapModeOverrides *
wrap_mode_overrides)
{
CoglHandle material;
const GList *layers;
int i;
GList *tmp;
GLfloat *v;
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
CoglMaterialFlushOptions options;
_COGL_GET_CONTEXT (ctx, NO_RETVAL);
material = ctx->source_material;
layers = cogl_material_get_layers (material);
/* Convert the vertices into an array of GLfloats ready to pass to
OpenGL */
for (v = (GLfloat *)ctx->logged_vertices->data, i = 0;
i < n_vertices;
v += stride, i++)
{
guint8 *c;
int j;
/* NB: [X,Y,Z,TX,TY...,R,G,B,A,...] */
v[0] = vertices[i].x;
v[1] = vertices[i].y;
v[2] = vertices[i].z;
for (tmp = (GList *)layers, j = 0; tmp != NULL; tmp = tmp->next, j++)
{
CoglHandle layer = (CoglHandle)tmp->data;
CoglHandle tex_handle;
GLfloat *t;
float tx, ty;
tex_handle = cogl_material_layer_get_texture (layer);
/* COGL_INVALID_HANDLE textures will be handled in
* _cogl_material_flush_layers_gl_state but there is no need to worry
* about scaling texture coordinates in this case */
if (tex_handle == COGL_INVALID_HANDLE)
continue;
tx = vertices[i].tx;
ty = vertices[i].ty;
_cogl_texture_transform_coords_to_gl (tex_handle, &tx, &ty);
/* NB: [X,Y,Z,TX,TY...,R,G,B,A,...] */
t = v + 3 + 2 * j;
t[0] = tx;
t[1] = ty;
}
if (use_color)
{
/* NB: [X,Y,Z,TX,TY...,R,G,B,A,...] */
c = (guint8 *) (v + 3 + 2 * n_layers);
c[0] = cogl_color_get_red_byte (&vertices[i].color);
c[1] = cogl_color_get_green_byte (&vertices[i].color);
c[2] = cogl_color_get_blue_byte (&vertices[i].color);
c[3] = cogl_color_get_alpha_byte (&vertices[i].color);
}
}
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
options.flags = COGL_MATERIAL_FLUSH_FALLBACK_MASK;
if (use_color)
options.flags |= COGL_MATERIAL_FLUSH_SKIP_GL_COLOR;
options.fallback_layers = fallback_layers;
if (wrap_mode_overrides)
{
options.flags |= COGL_MATERIAL_FLUSH_WRAP_MODE_OVERRIDES;
options.wrap_mode_overrides = *wrap_mode_overrides;
}
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
_cogl_material_flush_gl_state (ctx->source_material, &options);
GE (glDrawArrays (GL_TRIANGLE_FAN, 0, n_vertices));
}
void
cogl_polygon (const CoglTextureVertex *vertices,
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
unsigned int n_vertices,
gboolean use_color)
{
CoglHandle material;
const GList *layers, *tmp;
int n_layers;
gboolean use_sliced_polygon_fallback = FALSE;
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
guint32 fallback_layers = 0;
int i;
cogl: improves header and coding style consistency 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.
2010-02-09 20:57:32 -05:00
unsigned long enable_flags;
unsigned int stride;
gsize stride_bytes;
GLfloat *v;
CoglMaterialWrapModeOverrides wrap_mode_overrides;
CoglMaterialWrapModeOverrides *wrap_mode_overrides_p = NULL;
_COGL_GET_CONTEXT (ctx, NO_RETVAL);
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
_cogl_journal_flush ();
[draw-buffers] First pass at overhauling Cogl's framebuffer management Cogl's support for offscreen rendering was originally written just to support the clutter_texture_new_from_actor API and due to lack of documentation and several confusing - non orthogonal - side effects of using the API it wasn't really possible to use directly. This commit does a number of things: - It removes {gl,gles}/cogl-fbo.{c,h} and adds shared cogl-draw-buffer.{c,h} files instead which should be easier to maintain. - internally CoglFbo objects are now called CoglDrawBuffers. A CoglDrawBuffer is an abstract base class that is inherited from to implement CoglOnscreen and CoglOffscreen draw buffers. CoglOffscreen draw buffers will initially be used to support the cogl_offscreen_new_to_texture API, and CoglOnscreen draw buffers will start to be used internally to represent windows as we aim to migrate some of Clutter's backend code to Cogl. - It makes draw buffer objects the owners of the following state: - viewport - projection matrix stack - modelview matrix stack - clip state (This means when you switch between draw buffers you will automatically be switching to their associated viewport, matrix and clip state) Aside from hopefully making cogl_offscreen_new_to_texture be more useful short term by having simpler and well defined semantics for cogl_set_draw_buffer, as mentioned above this is the first step for a couple of other things: - Its a step toward moving ownership for windows down from Clutter backends into Cogl, by (internally at least) introducing the CoglOnscreen draw buffer. Note: the plan is that cogl_set_draw_buffer will accept on or offscreen draw buffer handles, and the "target" argument will become redundant since we will instead query the type of the given draw buffer handle. - Because we have a common type for on and offscreen framebuffers we can provide a unified API for framebuffer management. Things like: - blitting between buffers - managing ancillary buffers (e.g. attaching depth and stencil buffers) - size requisition - clearing
2009-09-25 09:34:34 -04:00
/* NB: _cogl_framebuffer_flush_state may disrupt various state (such
[draw-buffers] First pass at overhauling Cogl's framebuffer management Cogl's support for offscreen rendering was originally written just to support the clutter_texture_new_from_actor API and due to lack of documentation and several confusing - non orthogonal - side effects of using the API it wasn't really possible to use directly. This commit does a number of things: - It removes {gl,gles}/cogl-fbo.{c,h} and adds shared cogl-draw-buffer.{c,h} files instead which should be easier to maintain. - internally CoglFbo objects are now called CoglDrawBuffers. A CoglDrawBuffer is an abstract base class that is inherited from to implement CoglOnscreen and CoglOffscreen draw buffers. CoglOffscreen draw buffers will initially be used to support the cogl_offscreen_new_to_texture API, and CoglOnscreen draw buffers will start to be used internally to represent windows as we aim to migrate some of Clutter's backend code to Cogl. - It makes draw buffer objects the owners of the following state: - viewport - projection matrix stack - modelview matrix stack - clip state (This means when you switch between draw buffers you will automatically be switching to their associated viewport, matrix and clip state) Aside from hopefully making cogl_offscreen_new_to_texture be more useful short term by having simpler and well defined semantics for cogl_set_draw_buffer, as mentioned above this is the first step for a couple of other things: - Its a step toward moving ownership for windows down from Clutter backends into Cogl, by (internally at least) introducing the CoglOnscreen draw buffer. Note: the plan is that cogl_set_draw_buffer will accept on or offscreen draw buffer handles, and the "target" argument will become redundant since we will instead query the type of the given draw buffer handle. - Because we have a common type for on and offscreen framebuffers we can provide a unified API for framebuffer management. Things like: - blitting between buffers - managing ancillary buffers (e.g. attaching depth and stencil buffers) - size requisition - clearing
2009-09-25 09:34:34 -04:00
* as the material state) when flushing the clip stack, so should
* always be done first when preparing to draw. */
_cogl_framebuffer_flush_state (_cogl_get_framebuffer (), 0);
material = ctx->source_material;
layers = cogl_material_get_layers (ctx->source_material);
n_layers = g_list_length ((GList *)layers);
memset (&wrap_mode_overrides, 0, sizeof (wrap_mode_overrides));
for (tmp = layers, i = 0; tmp != NULL; tmp = tmp->next, i++)
{
CoglHandle layer = tmp->data;
CoglHandle tex_handle = cogl_material_layer_get_texture (layer);
/* COGL_INVALID_HANDLE textures will be handled in
* _cogl_material_flush_layers_gl_state */
if (tex_handle == COGL_INVALID_HANDLE)
continue;
2010-01-18 04:22:04 -05:00
/* Give the texture a chance to know that we're rendering
* non-quad shaped primitives. If the texture is in an atlas it
* will be migrated
*
* FIXME: this needs to be generalized. There could be any
* number of things that might require a shuffling of the
* underlying texture storage.
*/
2010-01-18 04:22:04 -05:00
_cogl_texture_ensure_non_quad_rendering (tex_handle);
/* We need to ensure the mipmaps are ready before deciding
* anything else about the texture because the texture storate
* could completely change if it needs to be migrated out of the
* atlas and will affect how we validate the layer.
*/
_cogl_material_layer_ensure_mipmaps (layer);
if (i == 0 && cogl_texture_is_sliced (tex_handle))
{
#if defined (HAVE_COGL_GLES) || defined (HAVE_COGL_GLES2)
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
g_warning ("cogl_polygon does not work for sliced textures "
"on GL ES");
warning_seen = TRUE;
return;
}
#endif
if (n_layers > 1)
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
{
g_warning ("Disabling layers 1..n since multi-texturing with "
"cogl_polygon isn't supported when using sliced "
"textures\n");
warning_seen = TRUE;
}
}
use_sliced_polygon_fallback = TRUE;
n_layers = 1;
[cogl] Move the texture filters to be a property of the material layer The texture filters are now a property of the material layer rather than the texture object. Whenever a texture is painted with a material it sets the filters on all of the GL textures in the Cogl texture. The filter is cached so that it won't be changed unnecessarily. The automatic mipmap generation has changed so that the mipmaps are only generated when the texture is painted instead of every time the data changes. Changing the texture sets a flag to mark that the mipmaps are dirty. This works better if the FBO extension is available because we can use glGenerateMipmap. If the extension is not available it will temporarily enable automatic mipmap generation and reupload the first pixel of each slice. This requires tracking the data for the first pixel. The COGL_TEXTURE_AUTO_MIPMAP flag has been replaced with COGL_TEXTURE_NO_AUTO_MIPMAP so that it will default to auto-mipmapping. The mipmap generation is now effectively free if you are not using a mipmap filter mode so you would only want to disable it if you had some special reason to generate your own mipmaps. ClutterTexture no longer has to store its own copy of the filter mode. Instead it stores it in the material and the property is directly set and read from that. This fixes problems with the filters getting out of sync when a cogl handle is set on the texture directly. It also avoids the mess of having to rerealize the texture if the filter quality changes to HIGH because Cogl will take of generating the mipmaps if needed.
2009-06-04 11:04:57 -04:00
if (cogl_material_layer_get_min_filter (layer) != GL_NEAREST
|| cogl_material_layer_get_mag_filter (layer) != GL_NEAREST)
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
{
g_warning ("cogl_texture_polygon does not work for sliced textures "
"when the minification and magnification filters are not "
"COGL_MATERIAL_FILTER_NEAREST");
warning_seen = TRUE;
}
return;
}
break;
}
if (cogl_texture_is_sliced (tex_handle))
{
static gboolean warning_seen = FALSE;
if (!warning_seen)
g_warning ("Disabling layer %d of the current source material, "
"because texturing with the vertex buffer API is not "
"currently supported using sliced textures, or "
"textures with waste\n", i);
warning_seen = TRUE;
[cogl] Improving Cogl journal to minimize driver overheads + GPU state changes Previously the journal was always flushed at the end of _cogl_rectangles_with_multitexture_coords, (i.e. the end of any cogl_rectangle* calls) but now we have broadened the potential for batching geometry. In ideal circumstances we will only flush once per scene. In summary the journal works like this: When you use any of the cogl_rectangle* APIs then nothing is emitted to the GPU at this point, we just log one or more quads into the journal. A journal entry consists of the quad coordinates, an associated material reference, and a modelview matrix. Ideally the journal only gets flushed once at the end of a scene, but in fact there are things to consider that may cause unwanted flushing, including: - modifying materials mid-scene This is because each quad in the journal has an associated material reference (i.e. not copy), so if you try and modify a material that is already referenced in the journal we force a flush first) NOTE: For now this means you should avoid using cogl_set_source_color() since that currently uses a single shared material. Later we should change it to use a pool of materials that is recycled when the journal is flushed. - modifying any state that isn't currently logged, such as depth, fog and backface culling enables. The first thing that happens when flushing, is to upload all the vertex data associated with the journal into a single VBO. We then go through a process of splitting up the journal into batches that have compatible state so they can be emitted to the GPU together. This is currently broken up into 3 levels so we can stagger the state changes: 1) we break the journal up according to changes in the number of material layers associated with logged quads. The number of layers in a material determines the stride of the associated vertices, so we have to update our vertex array offsets at this level. (i.e. calling gl{Vertex,Color},Pointer etc) 2) we further split batches up according to material compatability. (e.g. materials with different textures) We flush material state at this level. 3) Finally we split batches up according to modelview changes. At this level we update the modelview matrix and actually emit the actual draw command. This commit is largely about putting the initial design in-place; this will be followed by other changes that take advantage of the extended batching.
2009-06-17 13:46:42 -04:00
fallback_layers |= (1 << i);
continue;
}
/* By default COGL_MATERIAL_WRAP_MODE_AUTOMATIC becomes
GL_CLAMP_TO_EDGE but we want the polygon API to use GL_REPEAT
to maintain compatibility with previous releases */
if (cogl_material_layer_get_wrap_mode_s (layer) ==
COGL_MATERIAL_WRAP_MODE_AUTOMATIC)
{
wrap_mode_overrides.values[i].s =
COGL_MATERIAL_WRAP_MODE_OVERRIDE_REPEAT;
wrap_mode_overrides_p = &wrap_mode_overrides;
}
if (cogl_material_layer_get_wrap_mode_t (layer) ==
COGL_MATERIAL_WRAP_MODE_AUTOMATIC)
{
wrap_mode_overrides.values[i].t =
COGL_MATERIAL_WRAP_MODE_OVERRIDE_REPEAT;
wrap_mode_overrides_p = &wrap_mode_overrides;
}
}
/* Our data is arranged like:
* [X, Y, Z, TX0, TY0, TX1, TY1..., R, G, B, A,...] */
stride = 3 + (2 * n_layers) + (use_color ? 1 : 0);
stride_bytes = stride * sizeof (GLfloat);
/* Make sure there is enough space in the global vertex
array. This is used so we can render the polygon with a single
call to OpenGL but still support any number of vertices */
g_array_set_size (ctx->logged_vertices, n_vertices * stride);
v = (GLfloat *)ctx->logged_vertices->data;
/* Prepare GL state */
enable_flags = COGL_ENABLE_VERTEX_ARRAY;
enable_flags |= _cogl_material_get_cogl_enable_flags (ctx->source_material);
if (ctx->enable_backface_culling)
enable_flags |= COGL_ENABLE_BACKFACE_CULLING;
if (use_color)
{
enable_flags |= COGL_ENABLE_COLOR_ARRAY | COGL_ENABLE_BLEND;
GE( glColorPointer (4, GL_UNSIGNED_BYTE,
stride_bytes,
/* NB: [X,Y,Z,TX,TY...,R,G,B,A,...] */
v + 3 + 2 * n_layers) );
}
_cogl_enable (enable_flags);
_cogl_flush_face_winding ();
GE (glVertexPointer (3, GL_FLOAT, stride_bytes, v));
for (i = 0; i < n_layers; i++)
{
GE (glClientActiveTexture (GL_TEXTURE0 + i));
GE (glEnableClientState (GL_TEXTURE_COORD_ARRAY));
GE (glTexCoordPointer (2, GL_FLOAT,
stride_bytes,
/* NB: [X,Y,Z,TX,TY...,R,G,B,A,...] */
v + 3 + 2 * i));
}
_cogl_bitmask_clear_all (&ctx->temp_bitmask);
_cogl_bitmask_set_range (&ctx->temp_bitmask, n_layers, TRUE);
_cogl_disable_other_texcoord_arrays (&ctx->temp_bitmask);
if (use_sliced_polygon_fallback)
_cogl_texture_polygon_multiple_primitives (vertices,
n_vertices,
stride,
use_color);
else
_cogl_multitexture_polygon_single_primitive (vertices,
n_vertices,
n_layers,
stride,
use_color,
fallback_layers,
wrap_mode_overrides_p);
/* Reset the size of the logged vertex array because rendering
rectangles expects it to start at 0 */
g_array_set_size (ctx->logged_vertices, 0);
}