Source/platform/graphics/compositingThis directory contains the implementation of the “Blink compositing algorithm”.
This code is owned by the [paint team][paint-team-site]. [paint-team-site]: https://www.chromium.org/teams/paint-team
This document explains the SPv2 world as it develops, not the SPv1 world it replaces.
Design document: goo.gl/6xP8Oe
Inputs: PaintArtifact Outputs: List of cc::Layer objects and cc::PropertyTree's.
The algorithm walks through the list of PaintChunks in the PaintArtifact, allocating new c::Layers if the PaintChunk cannot merge into an existing cc::Layer. The reasons why it would not be able to do so are:
The PaintChunk requires a foreign layer (see below)
The PaintChunk cannot merge with any existing layer, due incompatible direct compositing reasons on its PropertyTreeState.
The PaintChunk overlaps with an earlier cc::Layer that it can't merge with due to reason 2, and there is no later-drawn cc::Layer for which reasons 1 and 2 do not apply.
In the worst case, this algorithm has an O(n^2) running time, where n is the number of PaintChunks.
All property tree nodes referred to by any PaintChunk are currently copied into their equivalent cc::PropertyTree node, regardless of whether they are required by the above.
When PaintChunks can merge into an existing cc::Layer, they may have different PropertyTreeStates than the PropertyTreeState of the cc::Layer. If so, we need to flatten down the nodes that are different between the PropertyTreeState of the PaintChunk and the cc::Layer. This is done by emitting paired display items to adjust the PaintChunks property tree state to the current state when the PaintChunk is consumed. See: ConversionContext::Convert.
Some PaintChunk content requires a foreign layer, meaning a layer that is managed outside of the scope of this code. Examples are composited video, a and 2D/3D (WebGL) canvas.
Any raster invalidates on a PaintChunk are also mapped to the space of the backing cc::Layer.