Currently the GL thread is started / stopped when the activity is respectively resumed / paused. However, according to the `GLSurfaceView` documentation, this should be done instead when the activity is started / stopped, so this change updates the start / stop logic for the GL thread to match the documentation.
`-Werror=array-bounds` flags when creating a new batch, possibly due to the possibility of the malloc failing (out of memory).
This PR adds an explicit `CRASHNOW` in the hope the compiler will recognise this case is not intended to be recoverable.
The architecture handling in 3.x is all over the place, and I'm not
going to start a major refactor like I did for 4.0 so late in its life
cycle.
So let's add more hacks! ⚔️
This requires manually passing `arch=arm64` when compiling, but this is
already the case for other parts of the codebase.
The commit b5a8055b5c should target GCC builds only as
-flax-vector-conversions has different behaviour in Clang and is
currently making the build fail.
Signed-off-by: Filipe Rinaldi <filipe.rinaldi@gmail.com>
(cherry picked from commit 2841144096)
This is a change done upstream in the `devel3` branch for 3.13.6:
82ca6b5ccb
They also seem to define it for macOS, but for us it breaks the build...
¯\_(ツ)_/¯
Also change wrong use of CPPFLAGS (pre-processor) where CXXFLAGS (C++)
makes more sense.
(cherry picked from commit b5a8055b5c)
- Implement shadow fading when using the Orthogonal shadow mode
(like in `master`).
This allows customizing the distance at which directional shadows
start to fade away. Shadow fading will also always start at the same
distance now, regardless of the current shadow mode in use.
This is useful for enclosed levels to prevent shadows from fading
at all with a well-tuned maximum distance.
The default fade start value (0.8) results in fading happening later
in the distance compared to the previous behavior, where fading started
from the last shadow split distance (0.6 in PSSM 4 Splits and
0.1 in PSSM 2 Splits).
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)
On some hardware, modifying gl_FragColor multiple times can cause large performance drops. This PR writes to a standard temporary variable instead, and copies across to gl_FragColor once only at the end of the fragment shader.
This could potentially lead to large gains in performance on affected hardware.
