On compulers that define __GNUC__ use computed goto to directly dispatch
the next instruction rather than going through another switch statement.
This saves a jump and some comparisons.
In tight loops this is is roughly 10% faster than the switch() method.
This implement branch prediction macros likely() and unlikely() like in
Linux. When using these macros please ensure that when you use them the
condition in the branch really is very, very likely or unlikely. Think
90+% of the time. Primarily useful for error checking. (And I implement
these macros for all our error checking macros now)
See this article for more information:
https://kernelnewbies.org/FAQ/LikelyUnlikely
There are more places where these macros may make sense in renderer and
physics engine. Placing them will come in another commit down the line.
We now allow booleanization of all types. This means that empty versions
of all types now evaluate to false. So a Vector2(0,0), Dictionary(),
etc.
This allows you to write GDScript like:
if not Dictionary():
print("Empty dict")
Booleanization can now also no longer fail. There is no more valid flag,
this changes Variant and GDNative API.
As a preparation for other performance enhancements to GDScript:call()
start by removing more of the GDScript runtime checks on release.
This code has been tested with 2d/platformer, 3d/platformer,
3d/materials_test, and goltorus. No regressions were found.
In an effort to make GDScript a little faster replace the double
switch() with a computed goto on compilers that set __GNUC__. For
compilers that don't support computed goto it will fall back to regular
switch/case statements.
In addition disable using boolean values in a mathematical context. Now
boolean values can only be compared with other booleans. Booleans will
also no longer be coerced to integers.
This PR replaces #11308 and fixes#11291
- Makes vararg methods automatically use PROPERTY_USAGE_NIL_IS_VARIANT on return types
- Completely removes the ":type" suffix for method names. Virtual methods must use the MethodInfo constructors that takes Variant::Type or PropertyHint as the first parameter for the return type (with CLASS_INFO as a helper to get the PropertyInfo). Parameters must use PROPERTY_HINT_RESOURCE_TYPE and hint string.
- PROPERTY_USAGE_NIL_IS_VARIANT is no longer needed for parameters, because parameters cannot be void.
- Adds missing PROPERTY_USAGE_NIL_IS_VARIANT to virtual and built-in methods that return Variant.
Currently we rely on some undefined behavior when Object->cast_to() gets
called with a Null pointer. This used to work fine with GCC < 6 but
newer versions of GCC remove all codepaths in which the this pointer is
Null. However, the non-static cast_to() was supposed to be null safe.
This patch makes cast_to() Null safe and removes the now redundant Null
checks where they existed.
It is explained in this article: https://www.viva64.com/en/b/0226/
The initial version of the pattern matcher in GDScript does not
allow matching on nested identifiers, only one identifiers available
in the current scope.
With the introduction of enums to GDScript that's a huge missing
feature. This commit makes the parser accept indexed constants and
variables to properly support enums.
GDScript was restricted to parse only scripts beginning with __res://__ or __user://__ to avoid templates from being parsed. I've made that a bit less inclusive by allowing scripts with an empty path to be parsed too, which doesn't conflict and is needed for this to work.
Also I've removed the `this` variable of the generated script and made the relevant object to be the one the script instance refers to, so you can use `self` instead.
Now, with the shorter 3.0-style syntax, you can write things like: `self.position.x + 10`
Closes#9500.
This removes `not` from the variable safe list of
keywords.
Before that this was a valid expression:
self.!(some_arg)
The other fix is just a forgotten boolean negation.
Notice: GDScript tokenizer used the old PoolFloatArray name.
Renamed PoolFloatArray to PoolRealArray.
Moved "project_settings.h" down one line to comply with the clang-format rules.
Fixes#9638
Closed pull request #9714 because I messed up with commits, sorry!
