This PR adds a define BVH_EXPAND_LEAF_AABBS which is set, which stores expanded AABBs in the tree instead of exact AABBs.
This makes the logic less error prone when considering reciprocal collisions in the pairing, as all collision detect is now taking place between expanded AABB against expanded AABB, rather than expanded AABB against exact AABB.
The flip side of this is that the intersection tests will now be less exact when expanded margins are set.
All margins are now user customizable via project settings, and take account of collision pairing density to adjust the margin dynamically.
Updating the broadphase to find new collision pairs was done after
checking for collision islands, so it was working in most cases due to
the pairing margin used in the BVH, but in case of teleported objects
the narrowphase collision could be skipped.
Now it's done before checking for collision islands, so we can ensure
that broadphase pairing has been done at the same time as objects are
marked as moved so their collision can be checked properly.
This issue didn't happen in the Octree/HashGrid because they do nothing
on update and trigger pairs directly when objects move instead.
Applying overlay materials into multi-surface meshes currently
requires adding a next pass material to all the surfaces, which
might be cumbersome when the material is to be applied to a range
of different geometries. This also makes it not trivial to use
AnimationPlayer to control the material in case of visual effects.
The material_override property is not an option as it works
replacing the active material for the surfaces, not adding a new pass.
This commit adds the material_overlay property to GeometryInstance
(and therefore MeshInstance), having the same reach as
material_override (that is, all surfaces) but adding a new material
pass on top of the active materials, instead of replacing them.
Implemented in rasterizer of both GLES2 and GLES3.
Correct transformation of normals that works with a Basis containing non-uniform scale is difficult to get correct for those not familiar with the maths, it is also rather verbose and hard to read in calling code. This PR adds helper functions which both standardize the approach and make it clearer in calling code what is being done and why.
Previously a crude metric was used to decide on the roaming expansion margin, but it created unexpected results in some scenarios. Instead this setting is exposed to the user via the RoomManager, allowing them to tailor it to the world size, room sizes, roaming objects sizes and the speeds of movement.
In all physics servers, body_get_direct_state() now silently returns
nullptr when the body has been already freed or is removed from space,
so the client code can detect this state and invalidate the body rid.
In 2D, there is no change in behavior (just no more errors).
In 3D, the Bullet server returned a valid direct body state when the
body was removed from the physics space, but in this case it didn't
make sense to use the information from the body state.
Async. compilation via ubershader is currently available in the scene and particles shaders only.
Bonus:
- Use `#if defined()` syntax for not true conditionals, so they don't unnecessarily take a bit in the version flagset.
- Remove unused `ENABLE_CLIP_ALPHA` from scene shader.
- Remove unused `PARTICLES_COPY` from the particles shader.
- Remove unused uniform related code.
- Shader language/compiler: use ordered hash maps for deterministic code generation (needed for caching).
Changes the Path2D drawing to use POLYLINE instead of thick lines.
Add a path to translate thick lines (that are not using anti-aliasing) to draw as polygons instead. This should be faster because polygons can be batched.
Sets `AlignOperands` to `DontAlign`.
`clang-format` developers seem to mostly care about space-based indentation and
every other version of clang-format breaks the bad mismatch of tabs and spaces
that it seems to use for operand alignment. So it's better without, so that it
respects our two-tabs `ContinuationIndentWidth`.
Changing the collision layer of a sleeping body was not triggering area
updates correctly.
Bodies need to be active for collision to be checked against already
overlapping bodies and areas.
Neighbors need to be activated too in order to handle the case where a
static body is modified (it can't be activated directly but paired
bodies need to check their collision again).
In 3D, moved the call to wakeup() from the physics server to
BodySW::_shapes_changed to make it consistent with 2D and also handle
the case where shapes are modified (_shapes_changed is called in both
this case and collision layer changes).
The BVH implementation is not checking collision layers on existing
pairs on move like other physics broadphases do.
This is solved by adding a new call to trigger pair callbacks again so
the physics engine can check layers again (specific to the BVH version,
other broadphase implementations just trigger a move like before).
This provides more realistic lighting with a very small performance cost.
The option is available in both GLES3 and GLES2, and can be enabled in
the Project Settings. This goes well with the ACES Fitted tonemapping mode
that was recently added.
When enabled, this also makes upgrading Godot 3.x projects to Godot 4.0 easier,
since lighting in 3.x will better match how it'll look in Godot 4.0.
These changes improve Rayshape behavior for Godot Physics 2D and 3D
when using move_and_slide with and without snapping.
Kinematic margin is now applied to ray shapes when handling snapping
collision tests and separation raycasts to help getting consistent
results in slopes and flat surfaces.
Recovery is calculated without the margin and a depth of 0 is still
considered a collision to stabilize results when on flat surface.
Recovery depth takes into account the current recovery vector (just like
test_body_motion) to fix jittering issues with multiple ray shapes due
to applying too much recovery.
Allows more flexible collision detection with different safe margin values.
Kinematic body motion changes in 2D and 3D:
-Recovery only for depth > min contact depth to help with collision
detection consistency (rest info could be lost if recovery was too much)
-Adaptive min contact depth (based on margin) instead of space parameter
Change the existing DEV_ASSERT function to be switched on and off by the DEV_ENABLED define. DEV_ASSERT breaks into the debugger as soon as hit.
Add error macros DEV_CHECK and DEV_CHECK_ONCE to add an alternative check that ERR_PRINT when a condition fails, again only enabled in DEV_ENABLED builds.
Browsers doesn't really like forcing the mix rate, e.g. Firefox does not
allow input (microphone) if the mix rate is not the default one, Chrom*
will exhibit worse performances, etc.
Sphere occluders are now tested for self occlusion. Spheres that are behind another sphere in the current view are superfluous so can be removed, cutting down on the runtime calculations.
AABBs are now maintained for Occluders as well as individual spheres, meaning a bunch of occluder spheres can be frustum rejected as a block.
Update mesh_surface_get_format_stride and
mesh_surface_make_offsets_from_format to return an array of offsets and
an array of strides in order to support vertex stream splitting
Update _get_array_from_surface to also support vertex stream splitting
Add a condition on split stream usage to ensure it does not get used on
dynamic meshes
Handle case when Tangent is compressed but Normal is not compressed
Make stream splitting option require a restart in the settings
Update SoftBody and Sprite3D to support and use strides and offsets
returned by updated visual_server functions
Update Sprite3D to use the dynamic mesh flag
Add framework for supporting geometrical occluders within rooms, and add support for sphere occluders.
Includes gizmos for editing.
They also work outside the portal system.
This was reported by UBSAN.
Many methods were discussed, in the end this has the least evils and will use a 0,0,1 default on decompress.
Please see the PR for more info https://github.com/godotengine/godot/pull/51268
This is only available on the GLES3 backend.
This can be useful for advanced shaders, but it should generally
not be enabled otherwise as full precision has a performance cost.
For general-purpose rendering, the built-in debanding filter should
be used to reduce banding instead.
Small bug in the logic, the roaming objects only should be set to done when they have been marked as visible, rather than the first time they are examined. This is because they can be seen in a room through multiple portals, and each needs to be tested until there is either a visible result or all the portals in are visited.
This backports the high quality glow mode from the `master` branch.
Previously, during downsample, every second row was ignored.
Now, when high-quality is used, we sample two rows at once to ensure
that no pixel is missed. It is slower, but looks much better and has
a much high stability while moving.
High quality also takes an additional horizontal sample the width of the
horizontal blur matches the height of the vertical blur.
Same thing that was already done in 2D, applies moving platform motion
by using a call to move_and_collide that excludes the platform itself,
instead of making it part of the body motion.
Helps with handling walls and slopes correctly when the character walks
on the moving platform.
Also made some minor adjustments to the 2D version and documentation.
Co-authored-by: fabriceci <fabricecipolla@gmail.com>
When synchronizing KinematicBody motion with moving the platform using
direct body state, only the linear velocity was taken into account.
This change exposes velocity at local point in direct body state and
uses it in move_and_slide to get the proper velocity that includes
rotations.
Fixed a bug in the complex PVS generation which was causing recursive loop.
Move some of the settings out of RoomManager into Project Settings.
Allow PVS generation method to be selected from Project Settings, and control PVS logging.
Fixes a bug whereby it read from the primary PVS in the gameplay monitor, using the size from the secondary PVS. This would read out of bounds and crash.
Removed debug code to update the gameplay monitor from the preview camera - this is no longer required.
Temporarily revert to the simple PVS generation method, because I've noticed a bug in the complex version, and the simple version is safer while I fix this.
With the octahedral compression, we had attributes of a size of 2 bytes
which potentially caused performance regressions on iOS/Mac
Now add padding to the normal/tangent buffer
For octahedral, normal will always be oct32 encoded
UNLESS tangent exists and is also compressed
then both will be oct16 encoded and packed into a vec4<GL_BYTE>
attribute
The existing tracing routine for building the PVS was rather simple compared to the main portal tracing, and could not correctly cope with paths that went through multiple portals from room A to B, and as a result would sometimes miss room entries in the PVS resulting in too many culled rooms in these circumstances.
This PR adds an improved function that can cope with entering a room multiple times during a trace. As a result it has to take care of portal directions (to prevent going back on itself) in a similar, but not identical way to the main portal tracing routine, and internal rooms, to prevent recursive loops.
The Transform::xform and xform_inv are made safe for Planes when using non-uniform scaling.
Basic unit tests for Transform.
Optimization of calling sites to prevent loss of performance from the changes to xform(Plane).
In some situations looking out from an internal room it was possible to look back into the portal into the internal room.
This PR fixes this by keeping a single item 'stack' record of the last external room, and preventing recursing into this room. This also makes tracing significantly more efficient out of internal rooms, as there is no need to trace the external room multiple times.
Initial octahedral compression incorrectly wrote tangent to the buffer
using an offset of 3 rather than 4, losing the sign of the tangent
vector needed for things like tangent space for texturing mapping
GLES3 renderer used remove_custom_define rather than set_conditional to
change back to the default conditional state the scene shader should be
in
In 3D, collision is disabled between kinematic/static bodies when
contacts are generated only to report them.
In 2D, this case was already fixed but the code is cleaned to make
it easier to follow.
This PR makes the 'convert rooms' button permanently on the toolbar and accessible whichever node is selected, so you can convert rooms without having to select the RoomManager first.
It also adds a togglable item 'view portal culling' to the 'View' menu which is a simple way of setting the RoomManager 'active' setting without the RoomManager being the selected node.
Both of these have keyboard shortcuts, which should make it much faster to reconvert rooms and edit.
In addition there the string in the 'Perspective' Listbox is modified to show [portals active] when portal culling is operational, for visual feedback. This is updated when you change modes, and when the rooms are invalidated.
When using the preview camera feature it turns out as well as culling the game objects, this also culls the editor gizmos from the preview camera, which makes the editor hard to use in this mode.
To get around this problem we simply disable frustum culling for GLOBAL portal_mode objects when in preview camera mode. This could be a bit slower in an editor scene with lots of gizmos but is the simplest way of solving the problem.
Implement Octahedral Compression for normal/tangent vectors
*Oct32 for uncompressed vectors
*Oct16 for compressed vectors
Reduces vertex size for each attribute by
*Uncompressed: 12 bytes, vec4<float32> -> vec2<unorm16>
*Compressed: 2 bytes, vec4<unorm8> -> vec2<unorm8>
Binormal sign is encoded in the y coordinate of the encoded tangent
Added conversion functions to go from octahedral mapping to cartesian
for normal and tangent vectors
sprite_3d and soft_body meshes write to their vertex buffer memory
directly and need to convert their normals and tangents to the new oct
format before writing
Created a new mesh flag to specify whether a mesh is using octahedral
compression or not
Updated documentation to discuss new flag/defaults
Created shader flags to specify whether octahedral or cartesian vectors
are being used
Updated importers to use octahedral representation as the default format
for importing meshes
Updated ShaderGLES2 to support 64 bit version codes as we hit the limit
of the 32-bit integer that was previously used as a bitset to store
enabled/disabled flags
Portal margins were not being correctly sent to the PortalRenderer from the SceneTree, so all margins were being used as default (1.0). This PR fixes this.
Implemented splitting of vertex positions and attributes in the vertex
buffer
Positions are sequential at the start of the buffer, followed by the
additional attributes which are interleaved
Made a project setting which enables/disabled the buffer formatting
throughout the project
Implemented in both GLES2 and GLES3
This improves performance particularly on tile-based GPUs as well as
cache performance for something like shadow mapping which only needs
position data
Updated Docs and Project Setting
* Safe and unsafe motion are calculated by dichotomy with a limited
number of steps. It's good for performance, but on long motions that
either collide near the beginning or near the end, the result can be
very imprecise.
* Now a factor 0.25 or 0.75 is used to converge faster when this case
happens, which allows longer motions to get more accurate collision
detection.
* Makes snap collision more precise, and helps with cases where diagonal collision on the border of a platform can lead to the character being stuck.
Additional improvements to move_and_slide:
* Handle slide canceling in move_and_collide with 0 velocity instead of
not applying it.
* Better handling of snap with custom logic to cancel sliding.
* Remove small jittering when using stop on slope, by canceling the
motion completely when the resulting motion is less than margin instead
of always projecting to the up direction (in both body motion and snap).
Co-authored-by: fabriceci <fabricecipolla@gmail.com>
Make sure the direction of the motion is preserved, unless the depth is
higher than the margin, which means the body needs depenetration in any
direction.
Also changed move_and_slide to avoid sliding on the first motion, in
order to avoid issues with unstable position on ground when jumping.
Co-authored-by: fabriceci <fabricecipolla@gmail.com>
It turned out the new autolinking feature was linking portals AFTER the static meshes had been added to rooms in the PortalRenderer. This meant that large meshes weren't being sprawled across these portals. The fix involves doing the autolinking BEFORE adding the static meshes.
Fixes a bug in the warning for portals being in the wrong direction, they should have only been checkout for outgoing portals. This was resulting in erroneous warnings.
Also the room conversion logs are refined to be more compact and informative.
A warning icon is also added in the gizmo for portals where autolink fails.
Fixing by applying the movement in two steps, first the platform
movement, and then the body movement. Plus, add the platform movement
when we are on_wall.
In 3D, disabled shapes are now not added to the broadphase anymore.
Since they are removed right away when disabled, no need to check for
disabled shapes for any query that comes from the broadphase.
Also Fixes raycast queries returning disabled shapes.
In 2D, disabled shapes where already not added to the broadphase.
Remove the same unnecessary checks as in 3D.
Overall harmonized API for disabled shapes in the physics servers and
removed duplicate method.
This is an older, easier to implement variant of CAS as a pure
fragment shader. It doesn't support upscaling, but we won't make
use of it (at least for now).
The sharpening intensity can be adjusted on a per-Viewport basis.
For the root viewport, it can be adjusted in the Project Settings.
Since `textureLodOffset()` isn't available in GLES2, there is no
way to support contrast-adaptive sharpening in GLES2.
It turns out the calculation of the bounding rect for canvas items has a nasty bug. When a transform is applied (especially in a custom draw), in the renderer this extra matrix is applied to all later commands in the canvas item. However in the calculation of the bound, the transform is only applied to the first command following the transform.
This PR fixes this inconsistency.
The code is based on the current version of thirdparty/vhacd and modified to use Godot's types and code style.
Additional changes:
- backported and extended PagedAllocator to allow leaked objects
- applied patch from https://github.com/bulletphysics/bullet3/pull/3037
Dynamic BVH doesn't update the tree anymore when calling set_pairable
with no parameter change.
Also modified Godot Physics broadphase to create objects directly with
pairable (static) set correctly to make use of this optimization for the
BVH broadphase.
Note: Octree broadphase doesn't use this optimization because it forces
an update on move, so passing the proper AABB and static parameters on
creation would cause the tree to update twice.
List of changes:
- Modified bvh class to handle 2D and 3D as a template
- Changes in Rect2, Vector2, Vector3 interface to uniformize template
calls
- New option in Project Settings to enable BVH for 2D Physics (enabled
by default like in 3D)
Since we clone the environments to build thirdparty code, we don't get an
explicit dependency on the build objects produced by that environment.
So when we update thirdparty code, Godot code using it is not necessarily
rebuilt (I think it is for changed headers, but not for changed .c/.cpp files),
which can lead to an invalid compilation output (linking old Godot .o files
with a newer, potentially ABI breaking version of thirdparty code).
This was only seen as really problematic with bullet updates (leading to
crashes when rebuilding Godot after a bullet update without cleaning .o files),
but it's safer to fix it everywhere, even if it's a LOT of hacky boilerplate.
(cherry picked from commit c7b53c03ae)
We've been using standard C library functions `memcpy`/`memset` for these since
2016 with 67f65f6639.
There was still the possibility for third-party platform ports to override the
definitions with a custom header, but this doesn't seem useful anymore.
Backport of #48239.
Allows users to override default API usage, in order to get best performance on different platforms.
Also changes the default legacy flags to use STREAM rather than DYNAMIC.
Fixes#46181
CameraServer.add_feed() takes a CameraFeed object type as parameter.
Passing in another type of data while binding the method it will make
tha parameter null.
Added a check for null which returns from function and does not make the
engine crash anymore.
(cherry picked from commit c158a63a8a)
The program would fail if the parameter is passed as null in set_primary_interface because
in the print_verbose, the get_namea) method is called on the parameter and this causes a
failure if the parameter that was passed is null.
The moment of inertia calculation for BoxShape is:
```
Vector3(
(p_mass / 3.0) * (ly * ly + lz * lz),
(p_mass / 3.0) * (lx * lx + lz * lz),
(p_mass / 3.0) * (lx * lx + ly * ly));
```
where the final line includes both the x and y extents.
However, for CapsuleShape3D, CylinderShape3D, ConvexPolygonShape3D, ConcavePolygonShape3D, and HeightMapShape3D, the final line read `(p_mass / 3.0) * (extents.y * extents.y + extents.y * extents.y)`. I believe this is a mistake, considering the comment in each case mentions using an AABB approximation, which should follow the same approach as BoxShape.
This change corrects the final line to include both the x and y components of the shape's extent.
The angular velocity estimate for kinematic bodies was calculated
incorrectly. Also, fixes its use in some kinematic/rigid collision
calculations.
3.3 version of #47130
Test specific axes before falling back to GJK-EPA algorithm to get more
accurate separation axes for common cases, the same way it's done for
cylinder-cylinder collision.
In the case of falling back to GJK-EPA algorithm to generate cylinder
contact points, margins were never taken into account.
This fixes the depenetration phase for kinematic bodies and allows
consistent floor detection for cylinder shapes.
- Fix objects with no material being considered as fully transparent by the lightmapper.
- Added "environment_min_light" property: gives artistic control over the shadow color.
- Fixed "Custom Color" environment mode, it was ignored before.
- Added "interior" property to BakedLightmapData: controls whether dynamic capture objects receive environment light or not.
- Automatically update dynamic capture objects when the capture data changes (also works for "energy" which used to require object movement to trigger the update).
- Added "use_in_baked_light" property to GridMap: controls whether the GridMap will be included in BakedLightmap bakes.
- Set "flush zero" and "denormal zero" mode for SSE2 instructions in the Embree raycaster. According to Embree docs it should give a performance improvement.
Changes default ninepatch mode to preserve compatibility, and renames default mode to 'fixed'.
Also adds an editor restart to changing ninepatch mode and software skinning, which will be more user friendly.
