Unity 性能优化

Unity版本为2022.3.34，本文所有的测试均在Editor下进行

尽管我认为，性能远远没有实现重要，提前优化、过度优化、虚空优化，之会影响开发效率，不过还是记录一下我最近学到的一些关于Unity性能优化的小技巧

脚本优化

测试工具

public class CustomTimer: IDisposable
{
    private string _name;
    private int _count;
    private Stopwatch _stopwatch;
    public CustomTimer(string name, int count)
    {
        _name = name;
        _count = count;
        if (_count <= 0)
        {
            _count = 1;
        }
        _stopwatch = Stopwatch.StartNew();
    }
    public void Dispose()
    {
        _stopwatch.Stop();
        UnityEngine.Debug.Log($"{_name} took {_stopwatch.ElapsedMilliseconds}ms for {_count.ToString("N0")} iterations, average time: {_stopwatch.ElapsedMilliseconds / _count}ms");
    }
}

GetComponent的方式

测试三种获得Component的方式

ComponentTest comp = null;
using (new CustomTimer("GetComponent(string)", testCount))
{
    for(int i = 0; i < testCount; i++)
    {
        comp = (ComponentTest)GetComponent("ComponentTest");
    }
}
using (new CustomTimer("GetComponent<>", testCount))
{
    for(int i = 0; i < testCount; i++)
    {
        comp = GetComponent<ComponentTest>();
    }
}
using (new CustomTimer("GetComponent(typeof(ComponentTest))", testCount))
{
    for(int i = 0; i < testCount; i++)
    {
        comp = (ComponentTest)GetComponent(typeof(ComponentTest));
    }
}

经测试，几乎无差别（貌似老版本会有区别，用string最慢）

GetComponent(string) took 136ms for 1,000,000 iterations, average time: 0ms
GetComponent<> took 115ms for 1,000,000 iterations, average time: 0ms
GetComponent(typeof(ComponentTest)) took 177ms for 1,000,000 iterations, average time: 0ms

比较go是否为null

int ans = 0;
using (new CustomTimer("Empty", testCount))
{
    for(int i = 0; i < testCount; i++)
    {
        ans++;
    }
}
Debug.Log(ans);
ans = 0;
using (new CustomTimer("!=", testCount))
{
    for(int i = 0; i < testCount; i++)
    {
        if (go != null)
        {
            ans++;
        }
    }
}
Debug.Log(ans);
ans = 0;
using (new CustomTimer("ReferenceEquals", testCount))
{
    for(int i = 0; i < testCount; i++)
    {
        if(!System.Object.ReferenceEquals(go, null))
        {
            ans++;
        }
    }
}
Debug.Log(ans);

经测试，直接比较是!= null更慢

Empty took 21ms for 100,000,000 iterations, average time: 0ms
!= took 445ms for 100,000,000 iterations, average time: 0ms
ReferenceEquals took 22ms for 100,000,000 iterations, average time: 0ms

检索Tag

int ans = 0;
using (new CustomTimer("string tag", testCount))
{
    for(int i = 0; i < testCount; i++)
    {
        if (go.tag == "Player")
        {
            ans++;
        }
    }
}
Debug.Log(ans);
ans = 0;
using (new CustomTimer("CompareTag", testCount))
{
    for(int i = 0; i < testCount; i++)
    {
        if (go.CompareTag("Player"))
        {
            ans++;
        }
    }
}
Debug.Log(ans);

直接比较tag字符串更慢

string tag took 162ms for 1,000,000 iterations, average time: 0ms
CompareTag took 70ms for 1,000,000 iterations, average time: 0ms

Update

停止远处GO的脚本update
分层update

序列化

重写序列化，只保留一些必须的参数，尽量存储简单的数据结构，反序列化时自行从Resources.LoadAsync()中加载出来

Unity 性能优化

Unity 性能优化

脚本优化

测试工具

GetComponent的方式

比较go是否为null

检索Tag

Update

序列化

图形优化

SRP Bather