This reverts commit f55039b194.
The GLES2 batching seems to require more testing and tweaking in order
to actually make the performance better on Android devices. It's been
proved with #21184 that the current implementation has it's drawbacks
therefore I suggest reverting the commit for now.
Adds GLES2 draw calls batching for the same render list item that uses
multiple rasterizer commands (e.g. Label node; a node with multiple
GDScript draw_* calls).
This commit unhacks some parts of the 3D rendering.
Most notably:
- possibility to use negative texture units
(no longer weird manual index allocation for user samplers)
- refactoring of light code, now sorts in a different way,
should yield better performance
- fixes a crash while saving (because of "Illegal instruction" execution)
when using a decent compiler (clang, it's clang. Thanks GCC for not telling me about UB).
Roughness is now set to 1 by default and albedo is now white,
even on meshes that do not have any materials defined.
This means there is no longer a visual difference between a
mesh with no materials defined and a mesh with a default
SpatialMaterial defined.
Shader compilation now keeps track of the discard key word.
Previously only variables were monitored. But discard, which needs
special treatment in some cases, went unnoticed by the compiler as
discard is not a variable but a flow control.
This commit adds monitoring for discard.
This commit makes operator[] on Vector const and adds a write proxy to it. From
now on writes to Vectors need to happen through the .write proxy. So for
instance:
Vector<int> vec;
vec.push_back(10);
std::cout << vec[0] << std::endl;
vec.write[0] = 20;
Failing to use the .write proxy will cause a compilation error.
In addition COWable datatypes can now embed a CowData pointer to their data.
This means that String, CharString, and VMap no longer use or derive from
Vector.
_ALWAYS_INLINE_ and _FORCE_INLINE_ are now equivalent for debug and non-debug
builds. This is a lot faster for Vector in the editor and while running tests.
The reason why this difference used to exist is because force-inlined methods
used to give a bad debugging experience. After extensive testing with modern
compilers this is no longer the case.
Note that gl_InstanceID is not supported in OpenGL ES 2.0,
so in the gles2 backend we assign it to 0.
Also clean up some duplicates/commented out code.
Fixes#20088.
