forked from lightspark/lightspark
-
Notifications
You must be signed in to change notification settings - Fork 0
/
README.GRAPHICS
30 lines (28 loc) · 1.76 KB
/
README.GRAPHICS
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Overview of graphics pipeline
-----------------------------
Lightspark tries to be efficient on the rendering side using both caching and
multi threaded rasterization.
Most of the graphic assets come in vectorial form. Cairo is used to transform
vectorial data in raster (bitmap format). PangoCairo is used for dynamic text.
Rendered data is then transferred onto OpenGL textures using Pixel Buffer
Objects (that hopefully are implemented using DMA). Textures are only loaded
when an object changes, so static objects are always cached in VRAM.
Pipeline for Stage:
1) When an object changes and needs to be redrawn it will call its
DisplayObject::requestInvalidation method. The method needs an InvalidateQueue
parameter and will add the object to such queue.
2) Invalidation requests are gathered together in the queue, but not executed
until the currently handled event ends, to make sure we are rendering a
consistent state.
3) At the end of the event all objects in the queue will have their ::invalidate
method called. This method returns an instance of IDrawable or NULL is rendering
is not necessary. It's important that the IDrawable instance contains a copy of
all the needed object state since the object may change while the rendering
happens. In the future Copy On Write could be useful to save the copy of data in
some cases.
4) The returned IDrawable and a reference to the object itself will then be
stored inside an AsyncDrawJob. The reference to the objects is necessary to make
sure the CachedSurface of the object, which stores information about the OpenGL
texture where the data must end, survives until the end of the rendering.
5) The AsyncDrawJob in added to the Thread Pool, to be executed. It will use
IDrawable::getPixelBuffer to get the raw pixels and upload them to the GPU.