Custom objects are slower to instantiate in Godot 4.0 GDScript compared to Godot 3.5.1

[MichaelReel]

Godot version

4.0 RC 3 4.0 RC 5

System information

Ubuntu 22.04.2 LTS, Intel® Core™ i7-8700K CPU @ 3.70GHz × 12, AMD® Radeon rx 6700 xt

Issue description

Apologies if this is a duplicate issue or a known feature of the new engine, I didn't discover anything with the same description in the bug tracker or the noted in the documentation.

I noticed that some code I ported from Godot 3.5 was significantly slower in Godot 4.0 RC 3. I whittled down the issue to the instantiation of a large number of custom defined objects that I was using.

Instantiating 10000 custom objects in Godot 3.5.1 might take around 15 milliseconds. Doing the same thing in Godot 4.0 RC 3 takes over a second on my hardware. The difference seems significant enough to me to wonder if this something that I will need to work around in Godot 4, or if it will be reverted in future.

Steps to reproduce

Writing a script in godot 3.5 that instantiates 10000 objects, custom objects (and for comparison custom resources), out puts some reasonable times:

Godot 3.5.1 Script:

class CustomObject extends Object:
	pass

class CustomResource extends Resource:
	pass

func _ready():
	var test_func_names = [
		"test_10000_objects",
		"test_10000_custom_objects",
		"test_10000_custom_resources",
	]
	for test_func_name in test_func_names:
		var test_func = funcref(self, test_func_name)
		var start_time: int = Time.get_ticks_usec()
		test_func.call_func()
		print("%s ran in %d usecs" % [test_func_name, Time.get_ticks_usec() - start_time])

func test_10000_objects():
	var test_array: Array = []
	for _i in range(10000):
		test_array.append(Object())

func test_10000_custom_objects():
	var test_array: Array = []
	for _i in range(10000):
		test_array.append(CustomObject.new())

func test_10000_custom_resources():
	var test_array: Array = []
	for _i in range(10000):
		test_array.append(CustomResource.new())

Godot 3.5.1 Output:

Godot Engine v3.5.1.stable.official.6fed1ffa3 - https://godotengine.org
OpenGL ES 3.0 Renderer: AMD Radeon RX 6700 XT (navi22, LLVM 15.0.6, DRM 3.47, 5.19.0-32-generic)
Async. shader compilation: OFF
 
test_10000_objects ran in 784 usecs
test_10000_custom_objects ran in 15344 usecs
test_10000_custom_resources ran in 25674 usecs

Running a similar script in Godot 4.0:

Godot 4.0 Script:

class CustomObject extends Object:
	pass

class CustomResource extends Resource:
	pass

func _ready():
	for test_func in [
		test_10000_objects,
		test_10000_custom_objects,
		test_10000_custom_resources,
	]:
		var start_time: int = Time.get_ticks_usec()
		test_func.call()
		print("%s ran in %d usecs" % [test_func.get_method(), Time.get_ticks_usec() - start_time])

func test_10000_objects():
	var test_array: Array = []
	for _i in range(10000):
		test_array.append(Object())

func test_10000_custom_objects():
	var test_array: Array = []
	for _i in range(10000):
		test_array.append(CustomObject.new())

func test_10000_custom_resources():
	var test_array: Array = []
	for _i in range(10000):
		test_array.append(CustomResource.new())

Godot 4.0 RC 5 Output:

Godot Engine v4.0.rc5.official.6296b4600 - https://godotengine.org
OpenGL API 4.6 (Core Profile) Mesa 22.2.5 - Compatibility - Using Device: AMD - AMD Radeon RX 6700 XT (navi22, LLVM 15.0.6, DRM 3.47, 5.19.0-32-generic)
 
test_10000_objects ran in 1435 usecs
test_10000_custom_objects ran in 1147367 usecs
test_10000_custom_resources ran in 12831 usecs

Minimal reproduction project

It's probably unnecessary, but I've included small projects that reproduce the issue with both Godot 4.0 RC 3 Forward+ and Compatibility set:

• InstancingTestG4RC3Compatibility.zip
• InstancingTestG4RC3Forward.zip

And I've also included a version of (more or less) the same code in Godot 3.5 as a comparison:

• InstancingTestG35OpenGLES3.zip

[MichaelReel]

I initially submitted this against RC3, but have just confirmed that I get the same results with RC5.