This fix works in both GLES3 and GLES2.
The rendering formula in the shader was adjusted to further improve the
sharpness/antialiasing quality balance.
- lawnjelly and Calinou
bc607fb607
Existing shadow caster culling using the BVH takes no account of the camera. This PR adds the highly encapsulated class VisualServerLightCuller which can cut down the casters in the shadow volume to only those which can cast shadows on the camera frustum.
This is used to:
* More accurately defer dirty updates to shadows when the shadow volume does not intersect the camera frustum.
* Tighter cull shadow casters to the view frustum.
Lights dirty state is now automatically managed:
* Continuous (tighter caster culling)
* Static (all casters are rendered)
- lawnjelly
8ca631a466
Allows a non-interpolated particle system to closely follow an interpolated target without tracking ahead of the target, by performing fixed timestep interpolation on the particle system global transform, and using this for emission.
- lawnjelly
3e19cf834a
Move VisualServer interpolation data out of Scenario and into VisualServerScene, so the interpolation data and enabled status is now common to all Scenarios.
Fix physics interpolation in multithreaded mode by ensuring tick and pre-draw are called.
- lawnjelly
d2b1d29634
Adds optional hierarchical culling to the 2D rendering (within VisualServer).
Each canvas item maintains a bound in local space of the item itself and all child / grandchild items. This allows branches to be culled at once when they don't intersect a viewport.
- lawnjelly
b777a9e5f9
It was insipred by godot 4's ProcessGroup system, however while Godot 4's implementation tries to hide threaded processing as much as possible, this implementation focuses on making it explicitly known and obvious to the user, in a (hopefully) almost bolierplate free way.
Also with the available options this node can be used for other purposes, like multi threaded cron job like method calls when paired with a Timer in manual mode.
* Node processing works on the concept of process groups.
* A node group can be inherited, run on main thread, or a sub-thread.
* Groups can be ordered.
* Process priority is now present for physics.
This is the first steps towards implementing godotengine/godot-proposals#6424.
No threading or thread guards exist yet in most of the scene code other than Node. That will have to be added later.
- reduz
98c655ec8d
- Only got the smaller improvements, and the thread safety for Node and SceneTree. I'm planning to implement a similar system, but I have a different way of doing it in mind.
Adds the option to change the audio driver to the Dummy driver and back at runtime, with a set of MuteState flags - Disabled (user control), Silence (period of silence), Focus Loss (when app is not in focus), and Paused (when app is paused).
Control for the flags is added for the editor in EditorSettings, and for the project in ProjectSettings.
Editor defaults to muted (Dummy driver) when there is no audio output, and automatically switches to active on output. This significantly reduces CPU usage.
- lawnjelly
cedb01fb84
* Adding and removing child nodes is now constant time, speed up should be huge.
* Searching for node paths as in ("path/to/node") should be far faster too.
This changes the children management and makes it a hashmap, optimizing most StringName based operations.
Most operations should be severe speed up without breaking compatibility.
This should fix many issues regarding to node access performance, and may also speed up editor start/end, but benchmarks are needed. So if you want to test, please make some benchmarks!
- reduz
ab5fc22f67
* This notification makes node children management very inefficient.
* Replaced by a NOTIFICATION_CHILDREN_CHANGED (and children_changed signal).
* Changed Canvas code (and similar) to use the above signal, to perform more efficiently.
This PR breaks compatibility (although this notification was very rarely used, even within the engine), but provides an alternate way to do the same.
It is required for the changes in #75627 to be entirely effective.
- reduz
Note that I removed NOTIFICATION_MOVED_IN_PARENT, as keeping it, but making it just not work is worse in my opinion.
104392ef4e
Calling queue_free() for large numbers of siblings could previously be very slow, with the time taken rising exponentially with number of children. This looked partly due to ordered_remove from the child list and notifications.
This PR identifies objects that are nodes, and sorts the deletion queue so that children are deleted in reverse child order. This minimizes the costs of reordering.
