GraphicsOutput

class GraphicsOutput

Bases: GraphicsOutputBase, DrawableRegion

This is a base class for the various different classes that represent the result of a frame of rendering. The most common kind of GraphicsOutput is a GraphicsWindow, which is a real-time window on the desktop, but another example is GraphicsBuffer, which is an offscreen buffer.

The actual rendering, and anything associated with the graphics context itself, is managed by the associated GraphicsStateGuardian (which might output to multiple GraphicsOutput objects).

GraphicsOutputs are not actually writable to bam files, of course, but they may be passed as event parameters, so they inherit from TypedWritableReferenceCount instead of TypedReferenceCount for that convenience.

Inheritance diagram

Inheritance diagram of GraphicsOutput

enum FrameMode

There are many reasons to call begin_frameend_frame.

enumerator FM_render = 0

We are rendering a frame.

enumerator FM_parasite = 1

We are rendering a frame of a parasite.

enumerator FM_refresh = 2

We are just refreshing the display or exposing the window.

enum RenderTextureMode
enumerator RTM_none = 0
enumerator RTM_bind_or_copy = 1

Try to render to the texture directly, but if that is not possible, fall back to RTM_copy_texture.

enumerator RTM_copy_texture = 2

Copy the image from the buffer to the texture every frame.

enumerator RTM_copy_ram = 3

Copy the image from the buffer to system RAM every frame.

enumerator RTM_triggered_copy_texture = 4

Copy the image from the buffer to the texture after a call to trigger_copy().

enumerator RTM_triggered_copy_ram = 5

Copy the image from the buffer to system RAM after a call to trigger_copy().

enumerator RTM_bind_layered = 6

Render directly to a layered texture, such as a cube map, 3D texture or 2D texture array. The layer that is being rendered to is selected by a geometry shader.

void add_render_texture(Texture *tex, RenderTextureMode mode, DrawableRegion::RenderTexturePlane bitplane = ::DrawableRegion::RTP_COUNT)

Creates a new Texture object, suitable for rendering the contents of this buffer into, and appends it to the list of render textures.

If tex is not NULL, it is the texture that will be set up for rendering into; otherwise, a new Texture object will be created, in which case you may call get_texture() to retrieve the new texture pointer.

You can specify a bitplane to attach the texture to. the legal choices are:

  • RTP_depth

  • RTP_depth_stencil

  • RTP_color

  • RTP_aux_rgba_0

  • RTP_aux_rgba_1

  • RTP_aux_rgba_2

  • RTP_aux_rgba_3

If you do not specify a bitplane to attach the texture to, this routine will use a default based on the texture’s format:

  • F_depth_component attaches to RTP_depth

  • F_depth_stencil attaches to RTP_depth_stencil

  • all other formats attach to RTP_color.

The texture’s format will be changed to match the format of the bitplane to which it is attached. For example, if you pass in an F_rgba texture and order that it be attached to RTP_depth_stencil, it will turn into an F_depth_stencil texture.

Also see make_texture_buffer(), which is a higher-level interface for preparing render-to-a-texture mode.

void clear_child_sort(void)

Resets the sort value of future offscreen buffers created by make_texture_sort() to the default value. See set_child_sort().

void clear_delete_flag(void)

Resets the delete flag, so the GraphicsOutput will not be automatically deleted before the beginning of the next frame.

void clear_render_textures(void)

If the GraphicsOutput is currently rendering to a texture, then all textures are dissociated from the GraphicsOuput.

int count_textures(void) const

If the GraphicsOutput is set to render into a texture, returns the number of textures that are being rendered into. Normally, the textures would be associated with different buffers - a color texture, a depth texture, and a stencil texture.

virtual bool flip_ready(void) const

These are not intended to be called directly by the user, but they’re published anyway since they might occasionally be useful for low-level debugging.

Returns true if a frame has been rendered and needs to be flipped, false otherwise.

PointerTo<DisplayRegion> get_active_display_region(int n) const

Returns the nth active DisplayRegion of those that have been created within the window. This may return NULL if n is out of bounds; particularly likely if the number of display regions has changed since the last call to get_num_active_display_regions().

int get_child_sort(void) const

Returns the sort value of future offscreen buffers created by make_texture_sort(). See set_child_sort().

static TypeHandle get_class_type(void)
bool get_delete_flag(void) const

Returns the current setting of the delete flag. When this is true, the GraphicsOutput will automatically be removed before the beginning of the next frame by the GraphicsEngine.

PointerTo<DisplayRegion> get_display_region(int n) const

Returns the nth DisplayRegion of those that have been created within the window. This may return NULL if n is out of bounds; particularly likely if the number of display regions has changed since the last call to get_num_display_regions().

GraphicsEngine *get_engine(void) const

Returns the graphics engine that created this output. Since there is normally only one GraphicsEngine object in an application, this is usually the same as the global GraphicsEngine.

FrameBufferProperties const &get_fb_properties(void) const

Returns the framebuffer properties of the window.

LVecBase2i get_fb_size(void) const

Returns the internal size of the window or buffer. This is almost always the same as get_size(), except when a pixel_zoom is in effect–see set_pixel_zoom().

int get_fb_x_size(void) const

Returns the internal width of the window or buffer. This is almost always the same as get_x_size(), except when a pixel_zoom is in effect–see set_pixel_zoom().

int get_fb_y_size(void) const

Returns the internal height of the window or buffer. This is almost always the same as get_y_size(), except when a pixel_zoom is in effect–see set_pixel_zoom().

GraphicsStateGuardian *get_gsg(void) const

Returns the GSG that is associated with this window. There is a one-to-one association between windows and GSG’s.

This may return NULL if the graphics context has not yet been created for the window, e.g. before the first frame has rendered; or after the window has been closed.

virtual GraphicsOutput *get_host(void)

This is normally called only from within make_texture_buffer(). When called on a ParasiteBuffer, it returns the host of that buffer; but when called on some other buffer, it returns the buffer itself.

bool get_inverted(void) const

Returns the current setting of the inverted flag. When this is true, the scene is rendered into the window upside-down, flipped like a mirror along the X axis. See set_inverted().

unsigned int get_left_eye_color_mask(void) const

Returns the color mask in effect when rendering a left-eye view in red_blue stereo mode. This is one or more bits defined in ColorWriteAttrib::Channels. See set_red_blue_stereo().

std::string const &get_name(void) const

Returns the name that was passed to the GraphicsOutput constructor.

int get_num_active_display_regions(void) const

Returns the number of active DisplayRegions that have been created within the window.

int get_num_display_regions(void) const

Returns the number of DisplayRegions that have been created within the window, active or otherwise.

bool get_one_shot(void) const

Returns the current setting of the one-shot flag. When this is true, the GraphicsOutput will automatically set itself inactive after the next frame.

DisplayRegion *get_overlay_display_region(void) const

Returns the special “overlay” DisplayRegion that is created for each window or buffer. This DisplayRegion covers the entire window, but cannot be used for rendering. It is a placeholder only, to indicate the dimensions of the window, and is usually used internally for purposes such as clearing the window, or grabbing a screenshot of the window.

There are very few applications that require access to this DisplayRegion. Normally, you should create your own DisplayRegion that covers the window, if you want to render to the window.

GraphicsPipe *get_pipe(void) const

Returns the GraphicsPipe that this window is associated with. It is possible that the GraphicsPipe might have been deleted while an outstanding PT(GraphicsOutput) prevented all of its children windows from also being deleted; in this unlikely case, get_pipe() may return NULL.

bool get_red_blue_stereo(void) const

Returns whether red-blue stereo mode is in effect for this particular window. See set_red_blue_stereo().

unsigned int get_right_eye_color_mask(void) const

Returns the color mask in effect when rendering a right-eye view in red_blue stereo mode. This is one or more bits defined in ColorWriteAttrib::Channels. See set_red_blue_stereo().

RenderTextureMode get_rtm_mode(int i = 0) const

Returns the RenderTextureMode associated with the nth render-texture. Returns RTM_none if there is no such texture.

LVecBase4 const &get_sbs_left_dimensions(void) const

Returns the effective sub-region of the window for displaying the left channel, if side-by-side stereo mode is in effect for the window. See set_side_by_side_stereo().

LVecBase2i get_sbs_left_size(void) const

If side-by-side stereo is enabled, this returns the pixel size of the left eye, based on scaling get_size() by get_sbs_left_dimensions(). If side-by- side stereo is not enabled, this returns the same as get_size().

int get_sbs_left_x_size(void) const

If side-by-side stereo is enabled, this returns the pixel width of the left eye, based on scaling get_x_size() by get_sbs_left_dimensions(). If side- by-side stereo is not enabled, this returns the same as get_x_size().

int get_sbs_left_y_size(void) const

If side-by-side stereo is enabled, this returns the pixel height of the left eye, based on scaling get_y_size() by get_sbs_left_dimensions(). If side-by-side stereo is not enabled, this returns the same as get_y_size().

LVecBase4 const &get_sbs_right_dimensions(void) const

Returns the effective sub-region of the window for displaying the right channel, if side-by-side stereo mode is in effect for the window. See set_side_by_side_stereo().

LVecBase2i get_sbs_right_size(void) const

If side-by-side stereo is enabled, this returns the pixel size of the right eye, based on scaling get_size() by get_sbs_right_dimensions(). If side- by-side stereo is not enabled, this returns the same as get_size().

int get_sbs_right_x_size(void) const

If side-by-side stereo is enabled, this returns the pixel width of the right eye, based on scaling get_x_size() by get_sbs_right_dimensions(). If side-by-side stereo is not enabled, this returns the same as get_x_size().

int get_sbs_right_y_size(void) const

If side-by-side stereo is enabled, this returns the pixel height of the right eye, based on scaling get_y_size() by get_sbs_right_dimensions(). If side-by-side stereo is not enabled, this returns the same as get_y_size().

bool get_screenshot(PNMImage &image)
PointerTo<Texture> get_screenshot(void)

Captures the most-recently rendered image from the framebuffer into the indicated PNMImage. Returns true on success, false on failure.

Captures the most-recently rendered image from the framebuffer and returns it as Texture, or NULL on failure.

bool get_side_by_side_stereo(void) const

Returns whether side-by-side stereo mode is in effect for this particular window. See set_side_by_side_stereo().

LVecBase2i const &get_size(void) const

Returns the visible size of the window or buffer, if it is known. In certain cases (e.g. fullscreen windows), the size may not be known until after the object has been fully created. Check has_size() first.

Certain objects (like windows) may change size spontaneously; this method is not thread-safe. To get the size of a window in a thread-safe manner, query get_properties().

int get_sort(void) const

Returns the sorting order of this particular GraphicsOutput. The various GraphicsOutputs within a particular thread will be rendered in the indicated order.

virtual bool get_supports_render_texture(void) const

Returns true if this particular GraphicsOutput can render directly into a texture, or false if it must always copy-to-texture at the end of each frame to achieve this effect.

bool get_swap_eyes(void) const

Returns the current setting of the “swap eyes” flag. See set_swap_eyes().

inline virtual Texture *get_texture(int i = 0) const

Returns the nth texture into which the GraphicsOutput renders. Returns NULL if there is no such texture.

If the texture is non-NULL, it may be applied to geometry to be rendered for any other windows or outputs that share the same GSG as this GraphicsOutput. The effect is undefined for windows that share a different GSG; usually in these cases the texture will be invalid.

NodePath get_texture_card(void)

Returns a PandaNode containing a square polygon. The dimensions are (-1,0,-1) to (1,0,1). The texture coordinates are such that the texture of this GraphicsOutput is aligned properly to the polygon. The GraphicsOutput promises to surgically update the Geom inside the PandaNode if necessary to maintain this invariant.

Each invocation of this function returns a freshly- allocated PandaNode. You can therefore safely modify the RenderAttribs of the PandaNode. The PandaNode is initially textured with the texture of this GraphicOutput.

DrawableRegion::RenderTexturePlane get_texture_plane(int i = 0) const

Returns the RenderTexturePlane associated with the nth render-texture. Returns 0 if there is no such texture.

int get_x_size(void) const

Returns the visible width of the window or buffer, if it is known. In certain cases (e.g. fullscreen windows), the size may not be known until after the object has been fully created. Check has_size() first.

Certain objects (like windows) may change size spontaneously; this method is not thread-safe. To get the size of a window in a thread-safe manner, query get_properties().

int get_y_size(void) const

Returns the visible height of the window or buffer, if it is known. In certain cases (e.g. fullscreen windows), the size may not be known until after the object has been fully created. Check has_size() first.

Certain objects (like windows) may change size spontaneously; this method is not thread-safe. To get the size of a window in a thread-safe manner, query get_properties().

bool has_size(void) const

Returns true if the size of the window/frame buffer is known, false otherwise. In certain cases the size may not be known until after the object has been fully created. Also, certain objects (like windows) may change size spontaneously.

bool has_texture(void) const

Returns true if the GraphicsOutput is rendering into any textures at all.

virtual bool is_active(void) const

Returns true if the window is ready to be rendered into, false otherwise.

bool is_nonzero_size(void) const

Returns true if the output has a nonzero size in both X and Y, or false if it is zero (and therefore invalid).

bool is_stereo(void) const

Returns Returns true if this window can render stereo DisplayRegions, either through red-blue stereo (see set_red_blue_stereo()) or through true hardware stereo rendering.

bool is_valid(void) const

Returns true if the output is fully created and ready for rendering, false otherwise.

GraphicsOutput *make_cube_map(std::string const &name, int size, NodePath &camera_rig, DrawMask camera_mask = PandaNode::get_all_camera_mask(), bool to_ram = false, FrameBufferProperties *fbp = nullptr)

This is similar to make_texture_buffer() in that it allocates a separate buffer suitable for rendering to a texture that can be assigned to geometry in this window, but in this case, the buffer is set up to render the six faces of a cube map.

The buffer is automatically set up with six display regions and six cameras, each of which are assigned the indicated draw_mask and parented to the given camera_rig node (which you should then put in your scene to render the cube map from the appropriate point of view).

You may take the texture associated with the buffer and apply it to geometry, particularly with TexGenAttrib::M_world_cube_map also in effect, to apply a reflection of everything seen by the camera rig.

DisplayRegion *make_display_region(void)
DisplayRegion *make_display_region(PN_stdfloat l, PN_stdfloat r, PN_stdfloat b, PN_stdfloat t)
DisplayRegion *make_display_region(LVecBase4 const &dimensions)

Creates a new DisplayRegion that covers the entire window.

If is_stereo() is true for this window, and default-stereo-camera is configured true, this actually makes a StereoDisplayRegion. Call make_mono_display_region() or make_stereo_display_region() if you want to insist on one or the other.

Creates a new DisplayRegion that covers the indicated sub-rectangle within the window. The range on all parameters is 0..1.

If is_stereo() is true for this window, and default-stereo-camera is configured true, this actually makes a StereoDisplayRegion. Call make_mono_display_region() or make_stereo_display_region() if you want to insist on one or the other.

Creates a new DisplayRegion that covers the indicated sub-rectangle within the window. The range on all parameters is 0..1.

If is_stereo() is true for this window, and default-stereo-camera is configured true, this actually makes a StereoDisplayRegion. Call make_mono_display_region() or make_stereo_display_region() if you want to insist on one or the other.

DisplayRegion *make_mono_display_region(void)
DisplayRegion *make_mono_display_region(PN_stdfloat l, PN_stdfloat r, PN_stdfloat b, PN_stdfloat t)
DisplayRegion *make_mono_display_region(LVecBase4 const &dimensions)

Creates a new DisplayRegion that covers the entire window.

This generally returns a mono DisplayRegion, even if is_stereo() is true. However, if side-by-side stereo is enabled, this will return a StereoDisplayRegion whose two eyes are both set to SC_mono. (This is necessary because in side-by-side stereo mode, it is necessary to draw even mono DisplayRegions twice).

Creates a new DisplayRegion that covers the entire window.

This generally returns a mono DisplayRegion, even if is_stereo() is true. However, if side-by-side stereo is enabled, this will return a StereoDisplayRegion whose two eyes are both set to SC_mono. (This is necessary because in side-by-side stereo mode, it is necessary to draw even mono DisplayRegions twice).

Creates a new DisplayRegion that covers the indicated sub-rectangle within the window. The range on all parameters is 0..1.

This generally returns a mono DisplayRegion, even if is_stereo() is true. However, if side-by-side stereo is enabled, this will return a StereoDisplayRegion whose two eyes are both set to SC_mono. (This is necessary because in side-by-side stereo mode, it is necessary to draw even mono DisplayRegions twice).

Filename make_screenshot_filename(std::string const &prefix = "screenshot")

Saves a screenshot of the region to a default filename, and returns the filename, or empty string if the screenshot failed. The default filename is generated from the supplied prefix and from the Config variable screenshot-filename, which contains the following strings:

%~p - the supplied prefix %~f - the frame count %~e - the value of screenshot-extension All other % strings in strftime().

StereoDisplayRegion *make_stereo_display_region(void)
StereoDisplayRegion *make_stereo_display_region(PN_stdfloat l, PN_stdfloat r, PN_stdfloat b, PN_stdfloat t)
StereoDisplayRegion *make_stereo_display_region(LVecBase4 const &dimensions)

Creates a new DisplayRegion that covers the entire window.

This always returns a stereo DisplayRegion, even if is_stereo() is false.

Creates a new DisplayRegion that covers the entire window.

This always returns a stereo DisplayRegion, even if is_stereo() is false.

Creates a new DisplayRegion that covers the indicated sub-rectangle within the window. The range on all parameters is 0..1.

This always returns a stereo DisplayRegion, even if is_stereo() is false.

GraphicsOutput *make_texture_buffer(std::string const &name, int x_size, int y_size, Texture *tex = nullptr, bool to_ram = false, FrameBufferProperties *fbp = nullptr)

Creates and returns an offscreen buffer for rendering into, the result of which will be a texture suitable for applying to geometry within the scene rendered into this window.

If you pass zero as the buffer size, the buffer will have the same size as the host window, and will automatically be resized when the host window is.

If tex is not NULL, it is the texture that will be set up for rendering into; otherwise, a new Texture object will be created. In either case, the target texture can be retrieved from the return value with buffer->get_texture() (assuming the return value is not NULL).

If to_ram is true, the buffer will be set up to download its contents to the system RAM memory associated with the Texture object, instead of keeping it strictly within texture memory; this is much slower, but it allows using the texture with any GSG.

This will attempt to be smart about maximizing render performance while minimizing framebuffer waste. It might return a GraphicsBuffer set to render directly into a texture, if possible; or it might return a ParasiteBuffer that renders into this window. The return value is NULL if the buffer could not be created for some reason.

When you are done using the buffer, you should remove it with a call to GraphicsEngine::remove_window().

void remove_all_display_regions(void)

Removes all display regions from the window, except the default one that is created with the window.

bool remove_display_region(DisplayRegion *display_region)

Removes the indicated DisplayRegion from the window, and destructs it if there are no other references.

Returns true if the DisplayRegion is found and removed, false if it was not a part of the window.

bool save_screenshot(Filename const &filename, std::string const &image_comment = "")

Saves a screenshot of the region to the indicated filename. The image comment is an optional user readable string that will be saved with the header of the image (if the file format supports embedded data; for example jpg allows comments). Returns true on success, false on failure.

Filename save_screenshot_default(std::string const &prefix = "screenshot")

Saves a screenshot of the region to a default filename, and returns the filename, or empty string if the screenshot failed. The filename is generated by make_screenshot_filename().

void set_active(bool active)

Sets the active flag associated with the GraphicsOutput. If the GraphicsOutput is marked inactive, nothing is rendered.

void set_child_sort(int child_sort)

Specifies the sort value of future offscreen buffers created by make_texture_sort().

The purpose of this method is to allow the user to limit the sort value chosen for a buffer created via make_texture_buffer(). Normally, this buffer will be assigned a value of get_sort() - 1, so that it will be rendered before this window is rendered; but sometimes this isn’t sufficiently early, especially if other buffers also have a view into the same scene.

If you specify a value here, then new buffers created via make_texture_buffer() will be given that sort value instead of get_sort() - 1.

void set_inverted(bool inverted)

Changes the current setting of the inverted flag. When this is true, the scene is rendered into the window upside-down and backwards, that is, inverted as if viewed through a mirror placed on the floor.

This is primarily intended to support DirectX (and a few buggy OpenGL graphics drivers) that perform a framebuffer-to-texture copy upside-down from the usual OpenGL (and Panda) convention. Panda will automatically set this flag for offscreen buffers on hardware that is known to do this, to compensate when rendering offscreen into a texture.

void set_one_shot(bool one_shot)

Changes the current setting of the one-shot flag. When this is true, the GraphicsOutput will render the current frame and then automatically set itself inactive. This is particularly useful for buffers that are created for the purposes of render-to-texture, for static textures that don’t need to be continually re-rendered once they have been rendered the first time.

Setting the buffer inactive is not the same thing as destroying it. You are still responsible for passing this buffer to GraphicsEngine::remove_window() when you no longer need the texture, in order to clean up fully. (However, you should not call remove_window() on this buffer while the texture is still needed, because depending on the render-to-texture mechanism in use, this may invalidate the texture contents.)

void set_overlay_display_region(DisplayRegion *display_region)

Replaces the special “overlay” DisplayRegion that is created for each window or buffer. See get_overlay_display_region(). This must be a new DisplayRegion that has already been created for this window, for instance via a call to make_mono_display_region(). You are responsible for ensuring that the new DisplayRegion covers the entire window. The previous overlay display region is not automatically removed; you must explicitly call remove_display_region() on it after replacing it with this method, if you wish it to be removed.

Normally, there is no reason to change the overlay DisplayRegion, so this method should be used only in very unusual circumstances.

void set_red_blue_stereo(bool red_blue_stereo, unsigned int left_eye_color_mask, unsigned int right_eye_color_mask)

Enables red-blue stereo mode on this particular window. When red-blue stereo mode is in effect, DisplayRegions that have the “left” channel set will render in the red (or specified) channel only, while DisplayRegions that have the “right” channel set will render in the blue (or specified) channel only.

The remaining two parameters specify the particular color channel(s) to associate with each eye. Use the bits defined in ColorWriteAttrib::Channels.

This can be used to achieve a cheesy stereo mode in the absence of hardware-supported stereo.

void set_side_by_side_stereo(bool side_by_side_stereo)
void set_side_by_side_stereo(bool side_by_side_stereo, LVecBase4 const &sbs_left_dimensions, LVecBase4 const &sbs_right_dimensions)

Enables side-by-side stereo mode on this particular window. When side-by- side stereo mode is in effect, DisplayRegions that have the “left” channel set will render on the part of the window specified by sbs_left_dimensions (typically the left half: (0, 0.5, 0, 1)), while DisplayRegions that have the “right” channel set will render on the part of the window specified by sbs_right_dimensions (typically the right half: (0.5, 1, 0, 1)).

This is commonly used in a dual-monitor mode, where a window is opened that spans two monitors, and each monitor represents a different eye.

virtual void set_sort(int sort)

Adjusts the sorting order of this particular GraphicsOutput, relative to other GraphicsOutputs.

void set_swap_eyes(bool swap_eyes)

Changes the “swap eyes” flag. This flag is normally false. When it is true, the left and right channels of a stereo DisplayRegion are sent to the opposite channels in the rendering backend. This is meant to work around hardware that inadvertently swaps the output channels, or hardware for which it cannot be determined which channel is which until runtime.

void setup_render_texture(Texture *tex, bool allow_bind, bool to_ram)

This is a deprecated interface that made sense back when GraphicsOutputs could only render into one texture at a time. From now on, use clear_render_textures and add_render_texture instead.

Deprecated: Use add_render_texture() instead.

virtual bool share_depth_buffer(GraphicsOutput *graphics_output)

Will attempt to use the depth buffer of the input graphics_output. The buffer sizes must be exactly the same.

AsyncFuture *trigger_copy(void)

When the GraphicsOutput is in triggered copy mode, this function triggers the copy (at the end of the next frame).

virtual void unshare_depth_buffer(void)

Discontinue sharing the depth buffer.