mutter/clutter/glx/clutter-stage-glx.c

749 lines
26 KiB
C
Raw Normal View History

2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
/* Clutter.
* An OpenGL based 'interactive canvas' library.
* Authored By Matthew Allum <mallum@openedhand.com>
* Copyright (C) 2006-2007 OpenedHand
*
* 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/>.
*
*
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "clutter-backend-glx.h"
#include "clutter-stage-glx.h"
#include "clutter-glx.h"
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
#include "clutter-profile.h"
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
#include "../clutter-main.h"
#include "../clutter-feature.h"
#include "../clutter-color.h"
#include "../clutter-util.h"
#include "../clutter-event.h"
#include "../clutter-enum-types.h"
#include "../clutter-private.h"
#include "../clutter-debug.h"
#include "../clutter-units.h"
#include "../clutter-shader.h"
#include "../clutter-group.h"
#include "../clutter-container.h"
#include "../clutter-stage.h"
#include "../clutter-stage-window.h"
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
#include "cogl/cogl.h"
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
#include <GL/glx.h>
#include <GL/gl.h>
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
#include <sys/stat.h>
#include <sys/ioctl.h>
#include <fcntl.h>
#include <errno.h>
#ifdef HAVE_DRM
#include <drm.h>
#endif
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
static void clutter_stage_window_iface_init (ClutterStageWindowIface *iface);
static ClutterStageWindowIface *clutter_stage_glx_parent_iface = NULL;
G_DEFINE_TYPE_WITH_CODE (ClutterStageGLX,
_clutter_stage_glx,
CLUTTER_TYPE_STAGE_X11,
G_IMPLEMENT_INTERFACE (CLUTTER_TYPE_STAGE_WINDOW,
clutter_stage_window_iface_init));
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
static void
clutter_stage_glx_unrealize (ClutterStageWindow *stage_window)
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
{
ClutterBackend *backend = clutter_get_default_backend ();
ClutterBackendX11 *backend_x11 = CLUTTER_BACKEND_X11 (backend);
ClutterStageX11 *stage_x11 = CLUTTER_STAGE_X11 (stage_window);
ClutterStageGLX *stage_glx = CLUTTER_STAGE_GLX (stage_window);
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
/* Note unrealize should free up any backend stage related resources */
CLUTTER_NOTE (BACKEND, "Unrealizing stage");
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
clutter_x11_trap_x_errors ();
if (stage_glx->glxwin != None)
{
glXDestroyWindow (backend_x11->xdpy, stage_glx->glxwin);
stage_glx->glxwin = None;
}
if (!stage_x11->is_foreign_xwin && stage_x11->xwin != None)
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
{
XDestroyWindow (backend_x11->xdpy, stage_x11->xwin);
stage_x11->xwin = None;
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
}
else
stage_x11->xwin = None;
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
XSync (backend_x11->xdpy, False);
clutter_x11_untrap_x_errors ();
CLUTTER_MARK ();
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
}
static gboolean
clutter_stage_glx_realize (ClutterStageWindow *stage_window)
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
{
ClutterStageX11 *stage_x11 = CLUTTER_STAGE_X11 (stage_window);
ClutterStageGLX *stage_glx = CLUTTER_STAGE_GLX (stage_window);
ClutterBackendX11 *backend_x11;
ClutterBackendGLX *backend_glx;
ClutterBackend *backend;
int event_flags;
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
2009-05-11 07:36:14 -04:00
CLUTTER_NOTE (ACTOR, "Realizing stage '%s' [%p]",
G_OBJECT_TYPE_NAME (stage_window),
stage_window);
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
backend = clutter_get_default_backend ();
backend_glx = CLUTTER_BACKEND_GLX (backend);
backend_x11 = CLUTTER_BACKEND_X11 (backend);
if (stage_x11->xwin == None)
{
XSetWindowAttributes xattr;
unsigned long mask;
XVisualInfo *xvisinfo;
gfloat width, height;
CLUTTER_NOTE (MISC, "Creating stage X window");
xvisinfo = clutter_backend_x11_get_visual_info (backend_x11);
if (xvisinfo == NULL)
{
g_critical ("Unable to find suitable GL visual.");
return FALSE;
}
/* window attributes */
xattr.background_pixel = WhitePixel (backend_x11->xdpy,
backend_x11->xscreen_num);
xattr.border_pixel = 0;
xattr.colormap = XCreateColormap (backend_x11->xdpy,
backend_x11->xwin_root,
xvisinfo->visual,
AllocNone);
mask = CWBorderPixel | CWColormap;
/* Call get_size - this will either get the geometry size (which
* before we create the window is set to 640x480), or if a size
* is set, it will get that. This lets you set a size on the
* stage before it's realized.
*/
clutter_actor_get_size (CLUTTER_ACTOR (stage_x11->wrapper),
&width,
&height);
stage_x11->xwin_width = (gint)width;
stage_x11->xwin_height = (gint)height;
stage_x11->xwin = XCreateWindow (backend_x11->xdpy,
backend_x11->xwin_root,
0, 0,
stage_x11->xwin_width,
stage_x11->xwin_height,
0,
xvisinfo->depth,
InputOutput,
xvisinfo->visual,
mask, &xattr);
CLUTTER_NOTE (BACKEND, "Stage [%p], window: 0x%x, size: %dx%d",
stage_window,
(unsigned int) stage_x11->xwin,
stage_x11->xwin_width,
stage_x11->xwin_height);
XFree (xvisinfo);
}
if (stage_glx->glxwin == None)
{
int major;
int minor;
GLXFBConfig config;
/* Try and create a GLXWindow to use with extensions dependent on
* GLX versions >= 1.3 that don't accept regular X Windows as GLX
* drawables. */
if (glXQueryVersion (backend_x11->xdpy, &major, &minor) &&
major == 1 && minor >= 3 &&
_clutter_backend_glx_get_fbconfig (backend_glx, &config))
{
stage_glx->glxwin = glXCreateWindow (backend_x11->xdpy,
config,
stage_x11->xwin,
NULL);
}
}
/* the masks for the events we want to select on a stage window;
* KeyPressMask and KeyReleaseMask are necessary even with XI1
* because key events are broken with that extension, and will
* be fixed by XI2
*/
event_flags = StructureNotifyMask
| FocusChangeMask
| ExposureMask
| PropertyChangeMask
| EnterWindowMask
| LeaveWindowMask
| KeyPressMask
| KeyReleaseMask;
/* if we don't use XI1 then we also want core pointer events */
if (!clutter_x11_has_xinput ())
event_flags |= (ButtonPressMask | ButtonReleaseMask | PointerMotionMask);
#ifdef HAVE_XINPUT
else
_clutter_x11_select_events (stage_x11->xwin);
#endif
/* we unconditionally select input events even with event retrieval
* disabled because we need to guarantee that the Clutter internal
* state is maintained when calling clutter_x11_handle_event() without
* requiring applications or embedding toolkits to select events
* themselves. if we did that, we'd have to document the events to be
* selected, and also update applications and embedding toolkits each
* time we added a new mask, or a new class of events.
*
* see: http://bugzilla.clutter-project.org/show_bug.cgi?id=998
* for the rationale of why we did conditional selection. it is now
* clear that a compositor should clear out the input region, since
* it cannot assume a perfectly clean slate coming from us.
*
* see: http://bugzilla.clutter-project.org/show_bug.cgi?id=2228
* for an example of things that break if we do conditional event
* selection.
*/
XSelectInput (backend_x11->xdpy, stage_x11->xwin, event_flags);
#ifdef GLX_INTEL_swap_event
if (clutter_feature_available (CLUTTER_FEATURE_SWAP_EVENTS))
{
GLXDrawable drawable =
stage_glx->glxwin ? stage_glx->glxwin : stage_x11->xwin;
/* similarly to above, we unconditionally select this event
* because we rely on it to advance the master clock, and
* drive redraw/relayout, animations and event handling.
*/
glXSelectEvent (backend_x11->xdpy,
drawable,
GLX_BUFFER_SWAP_COMPLETE_INTEL_MASK);
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
}
#endif /* GLX_INTEL_swap_event */
/* no user resize.. */
clutter_stage_x11_fix_window_size (stage_x11,
stage_x11->xwin_width,
stage_x11->xwin_height);
clutter_stage_x11_set_wm_protocols (stage_x11);
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
CLUTTER_NOTE (BACKEND, "Successfully realized stage");
/* chain up to the StageX11 implementation */
return clutter_stage_glx_parent_iface->realize (stage_window);
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
}
static int
clutter_stage_glx_get_pending_swaps (ClutterStageWindow *stage_window)
{
ClutterStageGLX *stage_glx = CLUTTER_STAGE_GLX (stage_window);
return stage_glx->pending_swaps;
}
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
static void
clutter_stage_glx_dispose (GObject *gobject)
{
G_OBJECT_CLASS (_clutter_stage_glx_parent_class)->dispose (gobject);
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
}
static void
_clutter_stage_glx_class_init (ClutterStageGLXClass *klass)
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
{
GObjectClass *gobject_class = G_OBJECT_CLASS (klass);
gobject_class->dispose = clutter_stage_glx_dispose;
}
static void
_clutter_stage_glx_init (ClutterStageGLX *stage)
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
{
}
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
static gboolean
clutter_stage_glx_has_redraw_clips (ClutterStageWindow *stage_window)
{
ClutterStageGLX *stage_glx = CLUTTER_STAGE_GLX (stage_window);
/* NB: a degenerate clip means a full stage redraw is required */
if (stage_glx->initialized_redraw_clip &&
stage_glx->bounding_redraw_clip.width != 0)
return TRUE;
else
return FALSE;
}
static gboolean
clutter_stage_glx_ignoring_redraw_clips (ClutterStageWindow *stage_window)
{
ClutterStageGLX *stage_glx = CLUTTER_STAGE_GLX (stage_window);
/* NB: a degenerate clip means a full stage redraw is required */
if (stage_glx->initialized_redraw_clip &&
stage_glx->bounding_redraw_clip.width == 0)
return TRUE;
else
return FALSE;
}
/* A redraw clip represents (in stage coordinates) the bounding box of
* something that needs to be redraw. Typically they are added to the
* StageWindow as a result of clutter_actor_queue_clipped_redraw() by
* actors such as ClutterGLXTexturePixmap. All redraw clips are
* discarded after the next paint.
*
* A NULL stage_clip means the whole stage needs to be redrawn.
*
* What we do with this information:
* - we keep track of the bounding box for all redraw clips
* - when we come to redraw; if the bounding box is smaller than the
* stage we scissor the redraw to that box and use
* GLX_MESA_copy_sub_buffer to present the redraw to the front
* buffer. Some heuristics are used to decide when a clipped redraw
* should be promoted into a full stage redraw.
*
* Currently we simply check that the bounding box height is < 300
* pixels.
*
* XXX: we don't have any empirical data telling us what a sensible
* thresholds is!
*
* TODO - we should use different heuristics depending on whether the
* framebuffer is on screen and not redirected by a compositor VS
* offscreen (either due to compositor redirection or because we are
* rendering to a CoglOffscreen framebuffer)
*
* When not redirected glXCopySubBuffer (on intel hardware at least)
* will block the GPU until the vertical trace is at the optimal point
* so the copy can be done without tearing. In this case we don't want
* to copy tall regions because they increase the average time spent
* blocking the GPU.
*
* When rendering offscreen (CoglOffscreen or redirected by
* compositor) then no extra synchronization is needed before the copy
* can start.
*
* In all cases we need to consider that glXCopySubBuffer implies a
* blit which may be avoided by promoting to a full stage redraw if:
* - the framebuffer is redirected offscreen or a CoglOffscreen.
* - the framebuffer is onscreen and fullscreen.
* By promoting to a full stage redraw we trade off the cost involved
* in rasterizing the extra pixels vs avoiding to use a blit to
* present the back buffer.
*
*/
static void
clutter_stage_glx_add_redraw_clip (ClutterStageWindow *stage_window,
ClutterGeometry *stage_clip)
{
ClutterStageGLX *stage_glx = CLUTTER_STAGE_GLX (stage_window);
/* If we are already forced to do a full stage redraw then bail early */
if (clutter_stage_glx_ignoring_redraw_clips (stage_window))
return;
/* A NULL stage clip means a full stage redraw has been queued and
* we keep track of this by setting a degenerate
* stage_glx->bounding_redraw_clip */
if (stage_clip == NULL)
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
{
stage_glx->bounding_redraw_clip.width = 0;
stage_glx->initialized_redraw_clip = TRUE;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
return;
}
/* Ignore requests to add degenerate/empty clip rectangles */
if (stage_clip->width == 0 || stage_clip->height == 0)
return;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
if (!stage_glx->initialized_redraw_clip)
{
stage_glx->bounding_redraw_clip.x = stage_clip->x;
stage_glx->bounding_redraw_clip.y = stage_clip->y;
stage_glx->bounding_redraw_clip.width = stage_clip->width;
stage_glx->bounding_redraw_clip.height = stage_clip->height;
}
else if (stage_glx->bounding_redraw_clip.width > 0)
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
{
clutter_geometry_union (&stage_glx->bounding_redraw_clip, stage_clip,
&stage_glx->bounding_redraw_clip);
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
}
#if 0
redraw_area = (stage_glx->bounding_redraw_clip.width *
stage_glx->bounding_redraw_clip.height);
stage_area = stage_x11->xwin_width * stage_x11->xwin_height;
/* Redrawing and blitting >70% of the stage is assumed to be more
* expensive than redrawing the additional 30% to avoid the blit.
*
* FIXME: This threshold was plucked out of thin air!
*/
if (redraw_area > (stage_area * 0.7f))
{
g_print ("DEBUG: clipped redraw too big, forcing full redraw\n");
/* Set a degenerate clip to force a full redraw */
stage_glx->bounding_redraw_clip.width = 0;
}
#endif
stage_glx->initialized_redraw_clip = TRUE;
}
static void
clutter_stage_window_iface_init (ClutterStageWindowIface *iface)
{
clutter_stage_glx_parent_iface = g_type_interface_peek_parent (iface);
iface->realize = clutter_stage_glx_realize;
iface->unrealize = clutter_stage_glx_unrealize;
iface->get_pending_swaps = clutter_stage_glx_get_pending_swaps;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
iface->add_redraw_clip = clutter_stage_glx_add_redraw_clip;
iface->has_redraw_clips = clutter_stage_glx_has_redraw_clips;
iface->ignoring_redraw_clips = clutter_stage_glx_ignoring_redraw_clips;
/* the rest is inherited from ClutterStageX11 */
2007-03-22 Emmanuele Bassi <ebassi@openedhand.com> * clutter/clutter-private.h: Remove inclusion of backend-specific headers; update the main context object; add the declarations for the event queue functions. * clutter/clutter-backend.[ch]: Add the abstract ClutterBackend object, which holds backend-specific settings, the main stage, and the event queue. Every backend must implement a subclass of ClutterBackend and ClutterStage. * clutter/clutter-feature.c: Protect the GLX specific calls behing #ifdef HAVE_CLUTTER_GLX. * clutter/clutter-actor.c: * clutter/clutter-group.c: * clutter/clutter-clone-texture.c: Include GL/gl.h * clutter/clutter-event.[ch]: Update public API and implement the event queue private API; hold a reference on the event objects; move out the keysym-to-unicode table; add the new event types. * clutter/clutter-color.h: Include clutter-fixed.h * clutter/clutter-main.c: Update API; get the main stage from the backend object; process the event received from the queue; lock/unlock the main mutex if we have one; move the initialisation process sooner in the init sequence, in order to have the backend object when we check for options; call the backed vfuncs in the pre/post parse hooks. * clutter/clutter-stage.c: Make ClutterStage and abstract class, implemented by the backends. * clutter/clutter/glx/clutter-glx.h: * clutter/clutter/glx/clutter-backend-glx.[ch]: * clutter/clutter/glx/clutter-event-glx.c: * clutter/clutter/glx/clutter-stage-glx.[ch]: * clutter/clutter/glx/Makefile.am: Add the GLX backend. * clutter/clutter/egl/clutter-backend-egl.[ch]: * clutter/clutter/egl/clutter-event-egl.c: * clutter/clutter/egl/clutter-stage-egl.[ch]: * clutter/clutter/egl/Makefile.am: Add the stub for a EGL backend. * examples/*.c: Update for the new API.
2007-03-22 14:21:59 -04:00
}
#ifdef HAVE_DRM
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
static int
drm_wait_vblank(int fd, drm_wait_vblank_t *vbl)
{
int ret, rc;
do
{
ret = ioctl(fd, DRM_IOCTL_WAIT_VBLANK, vbl);
vbl->request.type &= ~_DRM_VBLANK_RELATIVE;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
rc = errno;
}
while (ret && rc == EINTR);
return rc;
}
#endif /* HAVE_DRM */
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
static void
wait_for_vblank (ClutterBackendGLX *backend_glx)
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
{
if (backend_glx->vblank_type == CLUTTER_VBLANK_NONE)
return;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
if (backend_glx->wait_video_sync)
{
unsigned int retraceCount;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
CLUTTER_NOTE (BACKEND, "Waiting for vblank (wait_video_sync)");
backend_glx->get_video_sync (&retraceCount);
backend_glx->wait_video_sync (2,
(retraceCount + 1) % 2,
&retraceCount);
}
else
{
#ifdef HAVE_DRM
drm_wait_vblank_t blank;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
CLUTTER_NOTE (BACKEND, "Waiting for vblank (drm)");
blank.request.type = _DRM_VBLANK_RELATIVE;
blank.request.sequence = 1;
blank.request.signal = 0;
drm_wait_vblank (backend_glx->dri_fd, &blank);
#else
CLUTTER_NOTE (BACKEND, "No vblank mechanism found");
#endif /* HAVE_DRM */
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
}
}
void
_clutter_stage_glx_redraw (ClutterStageGLX *stage_glx,
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
ClutterStage *stage)
{
ClutterBackend *backend;
ClutterBackendX11 *backend_x11;
ClutterBackendGLX *backend_glx;
ClutterStageX11 *stage_x11;
GLXDrawable drawable;
unsigned int video_sync_count;
gboolean may_use_clipped_redraw;
gboolean use_clipped_redraw;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
CLUTTER_STATIC_TIMER (painting_timer,
"Redrawing", /* parent */
"Painting actors",
"The time spent painting actors",
0 /* no application private data */);
CLUTTER_STATIC_TIMER (swapbuffers_timer,
"Redrawing", /* parent */
"glXSwapBuffers",
"The time spent blocked by glXSwapBuffers",
0 /* no application private data */);
CLUTTER_STATIC_TIMER (blit_sub_buffer_timer,
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
"Redrawing", /* parent */
"glx_blit_sub_buffer",
"The time spent in _glx_blit_sub_buffer",
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
0 /* no application private data */);
backend = clutter_get_default_backend ();
backend_x11 = CLUTTER_BACKEND_X11 (backend);
backend_glx = CLUTTER_BACKEND_GLX (backend);
stage_x11 = CLUTTER_STAGE_X11 (stage_glx);
CLUTTER_TIMER_START (_clutter_uprof_context, painting_timer);
if (G_LIKELY (backend_glx->can_blit_sub_buffer) &&
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
/* NB: a degenerate redraw clip width == full stage redraw */
stage_glx->bounding_redraw_clip.width != 0 &&
/* some drivers struggle to get going and produce some junk
* frames when starting up... */
G_LIKELY (stage_glx->frame_count > 3) &&
/* While resizing a window clipped redraws are disabled to avoid
* artefacts. See clutter-event-x11.c:event_translate for a
* detailed explanation */
G_LIKELY (stage_x11->clipped_redraws_cool_off == 0))
may_use_clipped_redraw = TRUE;
else
may_use_clipped_redraw = FALSE;
if (may_use_clipped_redraw &&
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
G_LIKELY (!(clutter_paint_debug_flags &
CLUTTER_DEBUG_DISABLE_CLIPPED_REDRAWS)))
use_clipped_redraw = TRUE;
else
use_clipped_redraw = FALSE;
if (use_clipped_redraw)
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
{
cogl_clip_push_window_rectangle (stage_glx->bounding_redraw_clip.x,
stage_glx->bounding_redraw_clip.y,
stage_glx->bounding_redraw_clip.width,
stage_glx->bounding_redraw_clip.height);
_clutter_stage_do_paint (stage, &stage_glx->bounding_redraw_clip);
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
cogl_clip_pop ();
}
else
_clutter_stage_do_paint (stage, NULL);
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
if (clutter_paint_debug_flags & CLUTTER_DEBUG_REDRAWS &&
may_use_clipped_redraw)
{
ClutterGeometry *clip = &stage_glx->bounding_redraw_clip;
static CoglMaterial *outline = NULL;
CoglHandle vbo;
float x_1 = clip->x;
float x_2 = clip->x + clip->width;
float y_1 = clip->y;
float y_2 = clip->y + clip->height;
float quad[8] = {
x_1, y_1,
x_2, y_1,
x_2, y_2,
x_1, y_2
};
CoglMatrix modelview;
if (outline == NULL)
{
outline = cogl_material_new ();
cogl_material_set_color4ub (outline, 0xff, 0x00, 0x00, 0xff);
}
vbo = cogl_vertex_buffer_new (4);
cogl_vertex_buffer_add (vbo,
"gl_Vertex",
2, /* n_components */
COGL_ATTRIBUTE_TYPE_FLOAT,
FALSE, /* normalized */
0, /* stride */
quad);
cogl_vertex_buffer_submit (vbo);
cogl_push_matrix ();
cogl_matrix_init_identity (&modelview);
_clutter_actor_apply_modelview_transform (CLUTTER_ACTOR (stage),
&modelview);
cogl_set_modelview_matrix (&modelview);
cogl_set_source (outline);
cogl_vertex_buffer_draw (vbo, COGL_VERTICES_MODE_LINE_LOOP,
0 , 4);
cogl_pop_matrix ();
cogl_object_unref (vbo);
}
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
cogl_flush ();
CLUTTER_TIMER_STOP (_clutter_uprof_context, painting_timer);
if (stage_x11->xwin == None)
return;
drawable = stage_glx->glxwin ? stage_glx->glxwin : stage_x11->xwin;
/* If we might ever use _clutter_backend_glx_blit_sub_buffer then we
* always need to keep track of the video_sync_count so that we can
* throttle blits.
*
* Note: we get the count *before* we issue any glXCopySubBuffer or
* blit_sub_buffer request in case the count would go up before
* returning control to us.
*/
if (backend_glx->can_blit_sub_buffer && backend_glx->get_video_sync)
backend_glx->get_video_sync (&video_sync_count);
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
/* push on the screen */
if (use_clipped_redraw)
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
{
ClutterGeometry *clip = &stage_glx->bounding_redraw_clip;
ClutterGeometry copy_area;
CLUTTER_NOTE (BACKEND,
"_glx_blit_sub_buffer (window: 0x%lx, "
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
"x: %d, y: %d, "
"width: %d, height: %d)",
(unsigned long) drawable,
stage_glx->bounding_redraw_clip.x,
stage_glx->bounding_redraw_clip.y,
stage_glx->bounding_redraw_clip.width,
stage_glx->bounding_redraw_clip.height);
/* XXX: It seems there will be a race here in that the stage
* window may be resized before glXCopySubBufferMESA is handled
* and so we may copy the wrong region. I can't really see how
* we can handle this with the current state of X but at least
* in this case a full redraw should be queued by the resize
* anyway so it should only exhibit temporary artefacts.
*/
copy_area.y = clutter_actor_get_height (CLUTTER_ACTOR (stage))
- clip->y - clip->height;
copy_area.x = clip->x;
copy_area.width = clip->width;
copy_area.height = clip->height;
/* glXCopySubBufferMESA and glBlitFramebuffer are not integrated
* with the glXSwapIntervalSGI mechanism which we usually use to
* throttle the Clutter framerate to the vertical refresh and so
* we have to manually wait for the vblank period...
*/
/* Here 'is_synchronized' only means that the blit won't cause a
* tear, ie it won't prevent multiple blits per retrace if they
* can all be performed in the blanking period. If that's the
* case then we still want to use the vblank sync menchanism but
* we only need it to throttle redraws.
*/
if (!backend_glx->blit_sub_buffer_is_synchronized)
{
/* XXX: note that glXCopySubBuffer, at least for Intel, is
* synchronized with the vblank but glBlitFramebuffer may
* not be so we use the same scheme we do when calling
* glXSwapBuffers without the swap_control extension and
* call glFinish () before waiting for the vblank period.
*
* See where we call glXSwapBuffers for more details.
*/
glFinish ();
wait_for_vblank (CLUTTER_BACKEND_GLX (backend));
}
else if (backend_glx->get_video_sync)
{
/* If we have the GLX_SGI_video_sync extension then we can
* be a bit smarter about how we throttle blits by avoiding
* any waits if we can see that the video sync count has
* already progressed. */
if (backend_glx->last_video_sync_count == video_sync_count)
wait_for_vblank (CLUTTER_BACKEND_GLX (backend));
}
else
wait_for_vblank (CLUTTER_BACKEND_GLX (backend));
CLUTTER_TIMER_START (_clutter_uprof_context, blit_sub_buffer_timer);
_clutter_backend_glx_blit_sub_buffer (backend_glx,
drawable,
copy_area.x,
copy_area.y,
copy_area.width,
copy_area.height);
CLUTTER_TIMER_STOP (_clutter_uprof_context, blit_sub_buffer_timer);
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
}
else
{
CLUTTER_NOTE (BACKEND, "glXSwapBuffers (display: %p, window: 0x%lx)",
backend_x11->xdpy,
(unsigned long) drawable);
/* If we have GLX swap buffer events then glXSwapBuffers will return
* immediately and we need to track that there is a swap in
* progress... */
if (clutter_feature_available (CLUTTER_FEATURE_SWAP_EVENTS))
stage_glx->pending_swaps++;
if (backend_glx->vblank_type != CLUTTER_VBLANK_GLX_SWAP &&
backend_glx->vblank_type != CLUTTER_VBLANK_NONE)
{
/* If we are going to wait for VBLANK manually, we not only
* need to flush out pending drawing to the GPU before we
* sleep, we need to wait for it to finish. Otherwise, we
* may end up with the situation:
*
* - We finish drawing - GPU drawing continues
* - We go to sleep - GPU drawing continues
* VBLANK - We call glXSwapBuffers - GPU drawing continues
* - GPU drawing continues
* - Swap buffers happens
*
* Producing a tear. Calling glFinish() first will cause us
* to properly wait for the next VBLANK before we swap. This
* obviously does not happen when we use _GLX_SWAP and let
* the driver do the right thing
*/
glFinish ();
wait_for_vblank (CLUTTER_BACKEND_GLX (backend));
}
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
CLUTTER_TIMER_START (_clutter_uprof_context, swapbuffers_timer);
glXSwapBuffers (backend_x11->xdpy, drawable);
CLUTTER_TIMER_STOP (_clutter_uprof_context, swapbuffers_timer);
}
backend_glx->last_video_sync_count = video_sync_count;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
/* reset the redraw clipping for the next paint... */
stage_glx->initialized_redraw_clip = FALSE;
stage_glx->frame_count++;
Adds initial clipped redraw support to Clutter A new (internal only currently) API, _clutter_actor_queue_clipped_redraw can be used to queue a redraw along with a clip rectangle in actor coordinates. This clip rectangle propagates up to the stage and clutter backend which may optionally use the information to optimize stage redraws. The GLX backend in particular may scissor the next redraw to the clip rectangle and use GLX_MESA_copy_sub_buffer to present the stage subregion. The intention is that any actors that can naturally determine the bounds of updates should queue clipped redraws to reduce the cost of updating small regions of the screen. Notes: » If GLX_MESA_copy_sub_buffer isn't available then the GLX backend ignores any clip rectangles. » queuing multiple clipped redraws will result in the bounding box of each clip rectangle being used. » If a clipped redraw has a height > 300 pixels then it's promoted into a full stage redraw, so that the GPU doesn't end up blocking too long waiting for the vsync to reach the optimal position to avoid tearing. » Note: no empirical data was used to come up with this threshold so we may need to tune this. » Currently only ClutterX11TexturePixmap makes use of this new API. This is done via a new "queue-damage-redraw" signal that is emitted when the pixmap is updated. The default handler queues a clipped redraw with the assumption that the pixmap is being painted as a rectangle covering the actors transformed allocation. If you subclass ClutterX11TexturePixmap and change how it's painted you now also need to override the signal handler and queue your own redraw. Technically this is a semantic break, but it's assumed that no one is currently doing this. This still leaves a few unsolved issues with regards to optimizing sub stage redraws that need to be addressed in further work so this can only be considered a stepping stone a this point: » Because we have no reliable way to determine if the painting of any given actor is being modified any optimizations implemented using _clutter_actor_queue_redraw_with_clip must be overridable by a subclass, and technically must be opt-in for existing classes to avoid a change in semantics. E.g. consider that a user connects to the paint signal for ClutterTexture and paints a circle instead of a rectangle. In this case any original logic to queue clipped redraws would be incorrect. » Currently only the implementation of an actor has enough information with which to queue clipped redraws. E.g. It is not possible for generic code in clutter-actor.c to queue a clipped redraw when hiding an actor because actors have no way to report a "paint box". (remember actors can draw outside their allocation and actors with depth may also be projected outside of their allocation) » The current plan is to add a actor_class->get_paint_cuboid() virtual so actors can report a bounding cube for everything they would draw in their current state and use that to queue clipped redraws against the stage by projecting the paint cube into stage coordinates. » Our heuristics for promoting clipped redraws into full redraws to avoid blocking the GPU while we wait for the vsync need improving: » vsync issues aren't relevant for redirected/composited applications so they should use different heuristics. In this case we instead need to trade off the cost of blitting when using glXCopySubBuffer vs promoting to a full redraw and flipping instead.
2009-11-30 12:47:55 -05:00
}