Although the expanded bounds were working in normal use, for moving and growing objects, there was one case which was not dealt with properly - significant shrinkage of exact bounds within an expanded bound.
This PR detects significant shrinkage and forces a new expanded bound to be created.
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.
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`.
The pair user data wasn't correctly updated in the BVH from the check
pair callback.
This could lead to crashes when the check caused logical unpairing in
physics, then actual unpairing from the broadphase occured and the
physics server tried to delete an already deleted pointer.
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).
The previous error message incorrectly suggested that any Basis could be fixed by calling get_rotation_quation() or orthonormalize(). This PR points out that only a valid rotation Basis can be fixed in this way.
(cherry picked from commit d3a3b3aff3)
Added one more warning to the hideable warnings. These seem to be benign warnings and are hidden during use in rooms and portals. When used from other areas, only one warning is displayed per run, instead of for every occurrence.
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).
Clean: remove duplicate and interior vertices (uses Bullet algorithm)
Simplify: modify the geometry for further simplification (uses VHACD
algorithm)
In the editor, single convex hull now uses the clean option.
Added a new editor entry to create a simplified convex hull, can be
useful for creating convex hull from highly tessellated triangle meshes.
Specific change for 3.x:
Add support for Vector<Vector3> and PoolVector<Vector3> in the convex hull generator.
Changes passing of current_tree from a member variable to a function argument, making bugs due to stale state less likely.
Fix a bug in deactivate where current_tree variable was stale. This may have resulted in visual anomalies.
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
For the `master` branch, the minimum supported MSVC version is now
MSVC 2017 (with MSVC 2019 being recommended).
(cherry picked from commit b57d9c8005)
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)
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.
I had missed a special case check for userdata of two colliding objects being equal. In this case, collisions should not be reported. This is used in the physics to prevent self intersection in composite objects.
When making items visible from the visual server, the collision check is deferred to prevent two identical collision checks when set_pairable is called shortly after.
It turns out that for some items (especially meshes), set_pairable is never called. This PR detects this occurrence and forces a collision check at the end of the routine.
The input to smoothstep is not actually a weight, and the decscription
of smoothstep was pretty hard to understand and easy to misinterpret.
Clarified what it means to be approximately equal.
nearest_po2 does not do what the descriptions says it does. For one,
it returns the same power if the input is a power of 2. Second, it
returns 0 if the input is negative or 0, while the smallest possible
integral power of 2 actually is 1 (2^0 = 1). Due to the implementation
and how it is used in a lot of places, it does not seem wise to change
such a core function however, and I decided it is better to alter the
description of the built-in.
Added a few examples/clarifications/edge-cases.
(cherry picked from commit 7f9bfee0ac)
Channels that are inactive -or when playback has not started yet- will report -200 dB as their peak value (which is also the lowest value possible during playback).
(cherry picked from commit a2b3a73e2d)
A major feature lacking in the octree was proper support for setting visibility / activation. This meant that invisible objects were still causing lots of processing in the tree unnecessarily.
This PR adds proper support for activation, items are temporarily removed from the tree and collision detection when inactive.
Fix bug whereby AABBs were reused from previous items due to use of a pool, resulting in missed collisions.
Also use full mask collision checks for all cases except generic update.
In the octree collisions are flushed as objects are moved, whereas in the BVH they are usually flushed once per frame.
This was causing problems in the render tree in some rare situations where objects were being created (perhaps deleted and recreated using the same handle in the same frame). This PR flushes the collisions before creating objects, and set_pairable.
set_pairable may not be necessary but it is done for safety until proven not necessary.
Also a small potential for a bug is closed in remove_unordered use.
Change render BVH update scheme from once per update_dirty_instances to a new update_scenarios function called once per draw.
Fix lights not being properly unpaired.
Fixed bug in add_changed_item where AABBs were not being updated due to more than one update per tick.
Calling set_pairable now will update collisions (rather than waiting for the next item_move).
Mask behaviour used for pairing now (hopefully) matches octree.
Completely re-write the lightmap generation code:
- Follow the general lightmapper code structure from 4.0.
- Use proper path tracing to compute the global illumination.
- Use atlassing to merge all lightmaps into a single texture (done by @RandomShaper)
- Use OpenImageDenoiser to improve the generated lightmaps.
- Take into account alpha transparency in material textures.
- Allow baking environment lighting.
- Add bicubic lightmap filtering.
There is some minor compatibility breakage in some properties and methods
in BakedLightmap, but lightmaps generated in previous engine versions
should work fine out of the box.
The scene importer has been changed to generate `.unwrap_cache` files
next to the imported scene files. These files *SHOULD* be added to any
version control system as they guarantee there won't be differences when
re-importing the scene from other OSes or engine versions.
This work started as a Google Summer of Code project; Was later funded by IMVU for a good amount of progress;
Was then finished and polished by me on my free time.
Co-authored-by: Pedro J. Estébanez <pedrojrulez@gmail.com>
Happy new year to the wonderful Godot community!
2020 has been a tough year for most of us personally, but a good year for
Godot development nonetheless with a huge amount of work done towards Godot
4.0 and great improvements backported to the long-lived 3.2 branch.
We've had close to 400 contributors to engine code this year, authoring near
7,000 commit! (And that's only for the `master` branch and for the engine code,
there's a lot more when counting docs, demos and other first-party repos.)
Here's to a great year 2021 for all Godot users 🎆
(cherry picked from commit b5334d14f7)
Complete rewrite of spatial partitioning using a bounding volume hierarchy rather than octree.
Switchable in project settings between using octree or BVH for rendering and physics.
3.2 version of b5107715f1.
`get_seed()` still returns the previous state and not the initial seed,
because users may rely on this behavior for resetting the state in 3.2.
Documented this is going to be fixed in 4.0.
Co-authored-by: MidZik <matt.idzik1@gmail.com>
It's non-deterministic so it's better to show a fixed value like 0 instead of
having it potentially change whenever `randomize()` is called.
Fixes#43317.
(cherry picked from commit 35e6070a35)
Add __NetBSD__ to `platform_config.h` so that it can find `alloca`
and use the proper `pthread_setname_np` format.
Rename RANDOM_MAX to avoid conflict with NetBSD stdlib.
Fixes#42145.
(cherry picked from commit 5f4d64f4f3)
Prevents adding new octants until a limiting number of elements have been added to the current octant. This enables balancing the benefits of brute force against the benefits of spatial partitioning. The limit can be set per octree.
Project settings are added for rendering octree to set the best balance per project depending on number of tests per frame / tick, and the amount of editing of the octree.
Fixes octants being leaked when removing elements.
Optimize octree with cached linear lists
Storing elements in octants using linked lists is efficient for housekeeping but very slow for testing. This optimization stores additional local_vectors with Element pointers and AABBs which are cached and only updated when a dirty flag is set on the octant.
This is selectable with 2 versions of Octree : Octree and Octree_CL, Octree being the old behaviour. At present the cached list version is only used for the visual server octree (rendering) as it has only been demonstrated to be faster there so far.
This uses slightly more memory (probably a few kb in most cases) but can be significantly faster during testing (culling etc).
Co-authored-by: Sergey Minakov <naithar@icloud.com>
This reverts commit 7f61710183.
See #38868, in its current implementation a small skew value might end up
serialized to scene files due to floating point precision errors, which is
detrimental to VCS.
This can be cherry-picked anew once a fix for #38868 has been found.
- Fixed floating point issue on the old one.
- Fixed the equation on the get_euler_yxz function.
- Added unit tests.
This work has been kindly sponsored by IMVU.
(cherry picked from commit 2331300989)
Configured for a max line length of 120 characters.
psf/black is very opinionated and purposely doesn't leave much room for
configuration. The output is mostly OK so that should be fine for us,
but some things worth noting:
- Manually wrapped strings will be reflowed, so by using a line length
of 120 for the sake of preserving readability for our long command
calls, it also means that some manually wrapped strings are back on
the same line and should be manually merged again.
- Code generators using string concatenation extensively look awful,
since black puts each operand on a single line. We need to refactor
these generators to use more pythonic string formatting, for which
many options are available (`%`, `format` or f-strings).
- CI checks and a pre-commit hook will be added to ensure that future
buildsystem changes are well-formatted.
(cherry picked from commit cd4e46ee65)
Skew is x-axis only, because it must be bidirectionally convertible to a 2x3 matrix, but you can subtract it to the rotation to get the effect on y-axis
(cherry picked from commit efb1f7d76b)
This fixes numerous false positives coming out of the culling system.
AABB checks are now a full separating-axis check against the frustum, with the points of the frustum being compared to the planes of the box just as the points of the box were being compared to the planes of the frustum. This fixes large objects behind the camera not being culled correctly.
Some systems that used frustums that were (sometimes mistakenly?) unbounded on one or more side have been modified to be fully enclosed.
