Performance

Inherits: Object

Exposes performance-related data.

Description

This class provides access to a number of different monitors related to performance, such as memory usage, draw calls, and FPS. These are the same as the values displayed in the Monitor tab in the editor's Debugger panel. By using the get_monitor method of this class, you can access this data from your code.

You can add custom monitors using the add_custom_monitor method. Custom monitors are available in Monitor tab in the editor's Debugger panel together with built-in monitors.

Note: Some of the built-in monitors are only available in debug mode and will always return 0 when used in a project exported in release mode.

Note: Some of the built-in monitors are not updated in real-time for performance reasons, so there may be a delay of up to 1 second between changes.

Note: Custom monitors do not support negative values. Negative values are clamped to 0.

Methods

void

add_custom_monitor ( StringName id, Callable callable, Array arguments=[] )

Variant

get_custom_monitor ( StringName id )

StringName[]

get_custom_monitor_names ( )

float

get_monitor ( Monitor monitor ) const

int

get_monitor_modification_time ( )

bool

has_custom_monitor ( StringName id )

void

remove_custom_monitor ( StringName id )

Enumerations

enum Monitor:

  • TIME_FPS = 0 --- The number of frames rendered in the last second. This metric is only updated once per second, even if queried more often. Higher is better.

  • TIME_PROCESS = 1 --- Time it took to complete one frame, in seconds. Lower is better.

  • TIME_PHYSICS_PROCESS = 2 --- Time it took to complete one physics frame, in seconds. Lower is better.

  • MEMORY_STATIC = 3 --- Static memory currently used, in bytes. Not available in release builds. Lower is better.

  • MEMORY_STATIC_MAX = 4 --- Available static memory. Not available in release builds. Lower is better.

  • MEMORY_MESSAGE_BUFFER_MAX = 5 --- Largest amount of memory the message queue buffer has used, in bytes. The message queue is used for deferred functions calls and notifications. Lower is better.

  • OBJECT_COUNT = 6 --- Number of objects currently instantiated (including nodes). Lower is better.

  • OBJECT_RESOURCE_COUNT = 7 --- Number of resources currently used. Lower is better.

  • OBJECT_NODE_COUNT = 8 --- Number of nodes currently instantiated in the scene tree. This also includes the root node. Lower is better.

  • OBJECT_ORPHAN_NODE_COUNT = 9 --- Number of orphan nodes, i.e. nodes which are not parented to a node of the scene tree. Lower is better.

  • RENDER_TOTAL_OBJECTS_IN_FRAME = 10 --- The total number of objects in the last rendered frame. This metric doesn't include culled objects (either via hiding nodes, frustum culling or occlusion culling). Lower is better.

  • RENDER_TOTAL_PRIMITIVES_IN_FRAME = 11 --- The total number of vertices or indices rendered in the last rendered frame. This metric doesn't include primitives from culled objects (either via hiding nodes, frustum culling or occlusion culling). Due to the depth prepass and shadow passes, the number of primitives is always higher than the actual number of vertices in the scene (typically double or triple the original vertex count). Lower is better.

  • RENDER_TOTAL_DRAW_CALLS_IN_FRAME = 12 --- The total number of draw calls performed in the last rendered frame. This metric doesn't include culled objects (either via hiding nodes, frustum culling or occlusion culling), since they do not result in draw calls. Lower is better.

  • RENDER_VIDEO_MEM_USED = 13 --- The amount of video memory used (texture and vertex memory combined, in bytes). Since this metric also includes miscellaneous allocations, this value is always greater than the sum of RENDER_TEXTURE_MEM_USED and RENDER_BUFFER_MEM_USED. Lower is better.

  • RENDER_TEXTURE_MEM_USED = 14 --- The amount of texture memory used (in bytes). Lower is better.

  • RENDER_BUFFER_MEM_USED = 15 --- The amount of render buffer memory used (in bytes). Lower is better.

  • PHYSICS_2D_ACTIVE_OBJECTS = 16 --- Number of active RigidBody2D nodes in the game. Lower is better.

  • PHYSICS_2D_COLLISION_PAIRS = 17 --- Number of collision pairs in the 2D physics engine. Lower is better.

  • PHYSICS_2D_ISLAND_COUNT = 18 --- Number of islands in the 2D physics engine. Lower is better.

  • PHYSICS_3D_ACTIVE_OBJECTS = 19 --- Number of active RigidBody3D and VehicleBody3D nodes in the game. Lower is better.

  • PHYSICS_3D_COLLISION_PAIRS = 20 --- Number of collision pairs in the 3D physics engine. Lower is better.

  • PHYSICS_3D_ISLAND_COUNT = 21 --- Number of islands in the 3D physics engine. Lower is better.

  • AUDIO_OUTPUT_LATENCY = 22 --- Output latency of the AudioServer. Lower is better.

  • MONITOR_MAX = 23 --- Represents the size of the Monitor enum.

Method Descriptions

Adds a custom monitor with the name id. You can specify the category of the monitor using slash delimiters in id (for example: "Game/NumberOfNPCs"). If there is more than one slash delimiter, then the default category is used. The default category is "Custom". Prints an error if given id is already present.

func _ready():
    var monitor_value = Callable(self, "get_monitor_value")

    # Adds monitor with name "MyName" to category "MyCategory".
    Performance.add_custom_monitor("MyCategory/MyMonitor", monitor_value)

    # Adds monitor with name "MyName" to category "Custom".
    # Note: "MyCategory/MyMonitor" and "MyMonitor" have same name but different IDs, so the code is valid.
    Performance.add_custom_monitor("MyMonitor", monitor_value)

    # Adds monitor with name "MyName" to category "Custom".
    # Note: "MyMonitor" and "Custom/MyMonitor" have same name and same category but different IDs, so the code is valid.
    Performance.add_custom_monitor("Custom/MyMonitor", monitor_value)

    # Adds monitor with name "MyCategoryOne/MyCategoryTwo/MyMonitor" to category "Custom".
    Performance.add_custom_monitor("MyCategoryOne/MyCategoryTwo/MyMonitor", monitor_value)

func get_monitor_value():
    return randi() % 25

The debugger calls the callable to get the value of custom monitor. The callable must return a zero or positive integer or floating-point number.

Callables are called with arguments supplied in argument array.


Returns the value of custom monitor with given id. The callable is called to get the value of custom monitor. See also has_custom_monitor. Prints an error if the given id is absent.


Returns the names of active custom monitors in an Array.


Returns the value of one of the available built-in monitors. You should provide one of the Monitor constants as the argument, like this:

print(Performance.get_monitor(Performance.TIME_FPS)) # Prints the FPS to the console.

See get_custom_monitor to query custom performance monitors' values.


  • int get_monitor_modification_time ( )

Returns the last tick in which custom monitor was added/removed (in microseconds since the engine started). This is set to Time.get_ticks_usec when the monitor is updated.


Returns true if custom monitor with the given id is present, false otherwise.


Removes the custom monitor with given id. Prints an error if the given id is already absent.