Godot Performance Profiler

Overview

Analyzes Godot projects to detect performance bottlenecks in GDScript code. Identifies expensive operations in frame-critical functions like

_process

_physics_process

_draw

Core principle: Frame time is precious—expensive operations belong in

_ready

When to Use

Use when:

• Frame rate drops or inconsistent performance
• CPU usage is unexpectedly high
• Game stutters during gameplay

• _process

  functions contain complex logic
• Memory usage grows over time
• Before releasing a game or major update

Don't use for:

• Network/server performance issues (use server profiling tools)
• GPU/shader optimization (use Godot's GPU profiler)
• Physics simulation tuning (use Godot's physics debug tools)

Detection Patterns

Heavy _process Functions

Trigger: Functions with >15 lines of code in

_process

_physics_process

Detection:

# Count lines in process functions
rg "^func _process" -A 50 --glob "*.gd" | wc -l

Thresholds:

• Warning: >15 lines
• Critical: >30 lines

get_node() in Loops

Pattern:

# ❌ BAD: get_node() in _process
func _process(delta):
    get_node("UI/HealthBar").value = health  # Called every frame!
    get_node("Player").position = position

Detection regex:

func _process.*\n(?:.*\n)*?\s+get_node\(

Instantiation in _process

Pattern:

# ❌ BAD: Creating objects every frame
func _process(delta):
    var bullet = Bullet.new()  # Memory churn!
    add_child(bullet)

Detection keywords:

• .new()

  in

  _process

  or

  _physics_process

• .instantiate()

  in frame callbacks

• add_child()

  or

  remove_child()

  in loops

Complex Operations in _physics_process

Anti-patterns:

• Heavy pathfinding calculations
• Complex AI state machines
• Large array operations
• File I/O or network calls

Analysis Procedures

Frame Time Analysis

Steps:

1. Enable Godot profiler (Debug > Profiler)
2. Run game for 60 seconds of typical gameplay
3. Identify functions with high "Time (ms)" values
4. Sort by "Time %" to find top offenders
5. Look for functions >0.1ms per frame

Interpretation:

• <0.01ms: Excellent
• 0.01-0.05ms: Good
• 0.05-0.1ms: Acceptable

• 0.1ms: Needs optimization

Memory Usage Detection

Indicators:

• Continuous growth in Memory monitor
• Frequent garbage collection spikes
• Node count increasing over time

Detection:

# Check for object creation in process functions
rg "\.new\(\)|instantiate\(\)" -B 5 -A 2 --glob "*.gd" | \
  rg -A 10 "func _process|func _physics_process"

Draw Call Optimization

Check:

• Godot's "Monitors" tab > "Draw Calls"
• Each unique material = additional draw call
• GPU skinning vs CPU skinning

Optimization targets:

• <100 draw calls for 2D games
• <500 draw calls for simple 3D
• Use texture atlases to reduce material switches

Physics Performance

Red flags:

• High collision shape complexity
• Too many rigid bodies (>100)
• Complex polygon collisions

• _physics_process

  doing non-physics work

Optimization Suggestions

Cache Node References

Before:

extends CharacterBody2D

func _process(delta):
    get_node("UI/HealthBar").value = health
    get_node("UI/ManaBar").value = mana
    get_node("UI/LevelLabel").text = str(level)

After:

extends CharacterBody2D

@onready var health_bar = $UI/HealthBar
@onready var mana_bar = $UI/ManaBar
@onready var level_label = $UI/LevelLabel

func _process(delta):
    health_bar.value = health
    mana_bar.value = mana
    level_label.text = str(level)

Impact: Eliminates 3 node lookups per frame (~0.01ms each)

Move Initialization to _ready

Before:

func _process(delta):
    var gravity = ProjectSettings.get("physics/2d/default_gravity")
    velocity.y += gravity * delta

After:

var gravity

func _ready():
    gravity = ProjectSettings.get("physics/2d/default_gravity")

func _process(delta):
    velocity.y += gravity * delta

Object Pooling for Bullets/Particles

Before:

func shoot():
    var bullet = BulletScene.instantiate()
    bullet.position = global_position
    get_parent().add_child(bullet)

After:

var bullet_pool: Array[Bullet] = []

func _ready():
    # Pre-instantiate bullets
    for i in range(50):
        var bullet = BulletScene.instantiate()
        bullet.hide()
        bullet_pool.append(bullet)
        get_parent().add_child(bullet)

func shoot():
    for bullet in bullet_pool:
        if not bullet.visible:
            bullet.position = global_position
            bullet.show()
            bullet.activate()
            return

Impact: Eliminates instantiation overhead during gameplay

Signal-Based Updates

Before:

func _process(delta):
    # Checking every frame if health changed
    if health != previous_health:
        update_health_bar()
        previous_health = health

After:

signal health_changed(new_health)

var health = 100:
    set(value):
        if health != value:
            health = value
            health_changed.emit(health)

func _ready():
    health_changed.connect(update_health_bar)

Batch Array Operations

Before:

func _process(delta):
    for enemy in enemies:
        if enemy.position.distance_to(player.position) < 100:
            enemy.target_player()

After:

var check_timer = 0.0
const CHECK_INTERVAL = 0.1  # Check 10x per second, not 60x

func _process(delta):
    check_timer += delta
    if check_timer >= CHECK_INTERVAL:
        check_timer = 0
        update_enemy_targets()

func update_enemy_targets():
    for enemy in enemies:
        if enemy.position.distance_to(player.position) < 100:
            enemy.target_player()

Godot Profiler Integration

Enabling the Profiler

1. Run game with Debug > Start with Profiler
2. Or click the "Profiler" tab in the bottom panel while game runs
3. Enable specific monitors:
   • CPU Time
   • Function Time
   • Node Count
   • Memory
   • Draw Calls

Key Metrics to Monitor

Frame Time (ms):

• Shows total time per frame
• Target: <16.67ms for 60 FPS, <33.33ms for 30 FPS
• Spikes indicate hitches

Function Breakdown:

• Lists all functions sorted by time
• Look for

  _process

  ,

  _physics_process

  ,

  _draw

• Click function name to see callers

Memory Monitor:

• Watch for continuous growth
• Sudden spikes indicate allocations
• Plateaus followed by drops = garbage collection

Profiling Workflow

1. Establish baseline (unoptimized)
   └─ Record profiler data for 60 seconds

2. Identify top 3 time consumers
   └─ Sort by "Time %" in profiler

3. Apply optimization
   └─ Use patterns above

4. Validate improvement
   └─ Profile again, compare metrics
   └─ Verify frame time reduced

5. Repeat for next bottleneck

Examples

Example 1: UI Controller Optimization

Problem:

# ui_controller.gd
extends Control

func _process(delta):
    # Called every frame - 5 node lookups!
    get_node("HealthBar").value = player.health
    get_node("ManaBar").value = player.mana
    get_node("LevelLabel").text = "Level: " + str(player.level)
    get_node("XpBar").value = player.xp
    get_node("GoldLabel").text = "Gold: " + str(player.gold)

Profiler Output:

Function          | Time (ms) | Time %
----------------------------------------
_process          | 0.18      | 12.3%
get_node          | 0.15      | 10.2%  (x5)

Optimized:

# ui_controller.gd
extends Control

@onready var health_bar = $HealthBar
@onready var mana_bar = $ManaBar
@onready var level_label = $LevelLabel
@onready var xp_bar = $XpBar
@onready var gold_label = $GoldLabel

func _ready():
    # Connect to player signals instead of polling
    player.health_changed.connect(_on_health_changed)
    player.mana_changed.connect(_on_mana_changed)
    player.leveled_up.connect(_on_leveled_up)
    player.xp_changed.connect(_on_xp_changed)
    player.gold_changed.connect(_on_gold_changed)

func _on_health_changed(value): health_bar.value = value
func _on_mana_changed(value): mana_bar.value = value
func _on_leveled_up(level): level_label.text = "Level: " + str(level)
func _on_xp_changed(value): xp_bar.value = value
func _on_gold_changed(value): gold_label.text = "Gold: " + str(value)

Result:

Function          | Time (ms) | Time %
----------------------------------------
_process          | 0.00      | 0.0%   (removed!)
_on_health_changed| 0.01      | 0.7%   (event-driven)

Example 2: Enemy Spawner Fix

Problem:

# spawner.gd
extends Node2D

func _process(delta):
    if enemies.size() < max_enemies:
        var enemy = EnemyScene.instantiate()  # Memory churn!
        enemy.position = random_position()
        add_child(enemy)
        enemies.append(enemy)

Memory leak: Continuous instantiation without pooling

Optimized:

# spawner.gd
extends Node2D

var spawn_timer = 0.0
const SPAWN_RATE = 2.0  # Check spawn every 2 seconds

func _process(delta):
    spawn_timer += delta
    if spawn_timer >= SPAWN_RATE:
        spawn_timer = 0
        try_spawn()

func try_spawn():
    if enemies.size() < max_enemies:
        spawn_enemy()

func spawn_enemy():
    # Consider object pool for frequent spawns
    var enemy = EnemyScene.instantiate()
    enemy.position = random_position()
    add_child(enemy)
    enemies.append(enemy)

Additional improvement: Use object pooling for frequently spawned enemies

Example 3: Physics Optimization

Problem:

# player.gd
extends CharacterBody2D

func _physics_process(delta):
    # Heavy calculation every physics frame
    var nearby = get_tree().get_nodes_in_group("enemies")
    for enemy in nearby:
        if global_position.distance_to(enemy.global_position) < detection_radius:
            enemy.set_target(self)
    
    # Do physics
    velocity.y += gravity * delta
    move_and_slide()

Profiler Output:

Function              | Time (ms) | Time %
--------------------------------------------
_physics_process      | 0.45      | 28.5%
get_nodes_in_group    | 0.25      | 15.8%
distance_to           | 0.12      | 7.6%

Optimized:

# player.gd
extends CharacterBody2D

var detection_timer = 0.0
const DETECTION_RATE = 0.2  # 5x per second

func _physics_process(delta):
    # Separate physics from AI
    velocity.y += gravity * delta
    move_and_slide()
    
    # Run detection less frequently
    detection_timer += delta
    if detection_timer >= DETECTION_RATE:
        detection_timer = 0
        update_enemy_detection()

func update_enemy_detection():
    var nearby = get_tree().get_nodes_in_group("enemies")
    for enemy in nearby:
        if global_position.distance_to(enemy.global_position) < detection_radius:
            enemy.set_target(self)

Result: Physics frame time reduced by ~60%

Success Criteria

A performance optimization is successful when:

Quantitative Metrics

• Target function frame time reduced by >50%

• _process

  function line count <15 lines
• Zero

  get_node()

  calls in

  _process

  or

  _physics_process

• Zero

  .new()

  or

  .instantiate()

  calls in frame callbacks
• Profiler "Time %" for top 3 functions <30% combined
• Frame time stays <16.67ms for 60 FPS target
• Memory usage stable (no continuous growth)

Qualitative Checks

• Node references cached in

  _ready()

  or

  @onready

• Complex logic moved to signals or timers
• Object pooling used for frequent instantiations

• _physics_process

  contains only physics-related code
• Profiler data shows improvement vs baseline

Validation Steps

1. Profile before: Record baseline metrics
2. Apply optimization: Make targeted changes
3. Profile after: Compare metrics
4. Verify in-game: Test actual gameplay feels smoother
5. Check edge cases: Ensure optimization works in all scenarios

Quick Reference

get_node

_process

@onready

.new()

_physics_process

ProjectSettings.get()

_ready

Common Mistakes

Mistake: "I'll optimize later"

Problem: Technical debt accumulates, harder to fix later Fix: Profile early and often, fix bottlenecks as they appear

Mistake: Premature optimization

Problem: Optimizing code that isn't a bottleneck Fix: Always profile first, focus on top time consumers

Mistake: Micro-optimizations

Problem: Spending hours to save 0.001ms Fix: Target functions >0.1ms, ignore the rest

Mistake: Not validating improvements

Problem: Assuming optimization worked without measuring Fix: Always run profiler before and after

Mistake: Optimizing release builds only

Problem: Debug builds have different performance characteristics Fix: Profile release builds for accurate data