You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Collision for all regions currently builds when a scene runs using the physics server.
There is a debug/show_collision option that builds collision for all regions right then, in the editor and it also works in game. This creates a static body rather than the physics server. If view gizmos is on, you can see it in the editor. You also need this if you want to enable view collision shapes in game.
Collision takes up a lot of memory and is a bit slow to generate. Each region takes 145ms to generate on my system. And running 16k x 8k collision takes about 6gb, or about 48mb per region.
What I'd like is keep the two options of build all collision at runtime, build all collision in the editor, and add dynamic options that will build a much smaller collision mesh around the camera.
The code is already structured into Terrain3D::_build_collision, _update_collision, _destroy_collision, but it's not fully setup to be dynamic yet. So most of the work will be in the middle one. It currently builds a 1025x1025 collision mesh. 1025 because we need one more vertex to get 1024 squares.
Provide an option to set dynamic collision mesh size, defaulting to 65x65. I would give fixed options: 33, 65, 127. (Though you could call them 32, 64, 128 as that is the number of squares across). In the set size function, you'll have to destroy_collision and build_collision, if its different and appropriate for the mode (eg don't build in editor if an in game is specified).
In build, create the heightmapshape of the specified size.
Terrain3D::__process increase the distance_to a bit, 5-10 units, up to a maximum of 50% of clipmap_size so it calls snap less frequently.
Terrain3D::snap(), after it sets the position of the meshes, call update_collision to set the position and vertices of the heightmap shape.
collision/enabled and set_collision_enabled can be used to toggle it on or off in realitme. If collision isn't built and it's enabled, create collision. Otherwise if collision is built, just toggle it on or off on the physics server, calling PhysicsServer3D.body_set_shape_disabled(). Keep the current logic of if it's disabled, don't build collision.
In the collision properties for the inspector (Terrain3D::bind_methods) debug/show_collision can be moved in with the collision group and change to an enum for collision_mode: dynamic in game, dynamic in editor, full in game, full in editor, with the following meanings:
dynamic in game (new default, generate 65x65 around player in game w/ physics server)
dynamic in editor (generate 65x65 around player in editor & game w/ static body)
full in game (current default, build for all regions w/ physics server)
full in editor (current debug/show_collision, build for all regions in editor & game w/ a static body)
The text was updated successfully, but these errors were encountered:
I updated the plan for collision options above. set_collision_enabled should be something toggleable, which means it can just talk to the physics server in realtime. Currently it is used only to determine if collision gets built.
Collision for all regions currently builds when a scene runs using the physics server.
There is a debug/show_collision option that builds collision for all regions right then, in the editor and it also works in game. This creates a static body rather than the physics server. If view gizmos is on, you can see it in the editor. You also need this if you want to enable view collision shapes in game.
Collision takes up a lot of memory and is a bit slow to generate. Each region takes 145ms to generate on my system. And running 16k x 8k collision takes about 6gb, or about 48mb per region.
What I'd like is keep the two options of build all collision at runtime, build all collision in the editor, and add dynamic options that will build a much smaller collision mesh around the camera.
The code is already structured into
Terrain3D::_build_collision
,_update_collision
,_destroy_collision
, but it's not fully setup to be dynamic yet. So most of the work will be in the middle one. It currently builds a 1025x1025 collision mesh. 1025 because we need one more vertex to get 1024 squares.Terrain3D::__process
increase the distance_to a bit, 5-10 units, up to a maximum of 50% ofclipmap_size
so it calls snap less frequently.Terrain3D::snap()
, after it sets the position of the meshes, call update_collision to set the position and vertices of the heightmap shape.collision/enabled
andset_collision_enabled
can be used to toggle it on or off in realitme. If collision isn't built and it's enabled, create collision. Otherwise if collision is built, just toggle it on or off on the physics server, callingPhysicsServer3D.body_set_shape_disabled()
. Keep the current logic of if it's disabled, don't build collision.Terrain3D::bind_methods
)debug/show_collision
can be moved in with the collision group and change to an enum for collision_mode: dynamic in game, dynamic in editor, full in game, full in editor, with the following meanings:The text was updated successfully, but these errors were encountered: