Unity performance optimization scheme arrangement

Inclusion size optimization

• Android ： Update the compression format of opacity map to ETC 4bit, because android Mobile phones in the market GPU There are many kinds of ,
  Every family's GPU Supports different compression formats , But they are compatible ETC Format , There are transparent channels ETC2 8bit
• Map Compression ： Require art to use 16 Bit mapping , We choose rgba16 position , Compress
• For transparent maps , We can only choose RGBA 16bit perhaps RGBA 32bit

CPU performance optimization

• Reduce FPS, stay ProjectSetting-> Quality Medium VSync Count Parameters will affect your FPS,EveryVBlank amount to FPS=60,EverySecondVBlank = 30; Neither of these situations is consistent with the game FPS Words , We need to manually adjust FPS, First, turn off the function of vertical synchronization , Then in the code Awake Set manually in the method FPS（Application.targetFrameRate = 45;）
  Reduce FPS The benefits of ：
  1） Save electricity , Reduce the heating of mobile phones ;
  2） Can stabilize the game FPS, Reduce the occurrence of jamming .

• When we set up FPS after , Adjust again Fixed timestep This parameter ,
  This parameter is in ProjectSetting->Time in , The purpose is to reduce the number of physical calculations , To improve game performance .

• Use... As little as possible Update LateUpdate FixedUpdate, This can also improve performance and save power .
  Use more events （ His writing , perhaps C# Event delegation in ）.

• For the vertices of the model （<500）, Number of faces （300-1500）, Map size (512*512) To standardize （ It was formulated when the project was approved ）

• Number of light sources ： Suggest 1 A directional light , The more light sources ,drawcall The more

• Try to mark non moving objects Static label ,Unity At run time static Automatic optimization of objects , Therefore, non operating entities should be checked as much as possible static label .

• OnBecameInvisible() When it becomes invisible and OnBecameVisible() When it becomes visible .
  Take this 2 The premise for a function to work is , The GameObject The binding MeshRender Components .

Memory optimization

• Be sure to delete the default methods that are empty or unnecessary in the script ;
• Use only in one script OnGUI Method ;
• To avoid the OnGUI Of variables in 、 Method to update 、 assignment , Output variables are recommended in Update Inside ;
• For variables frequently used in the same script, it is recommended to declare them as global variables , Variables or methods frequently called between scripts are recommended to be declared as global static variables or methods ;
• Don't get components frequently , Declare it as a global variable ;
• Array 、 Collection elements are preferred Array, The second is List;
• Scripts are disabled when not in use , Enable when necessary ;
• have access to Ray Instead of OnMouseXXX Class method ;
• Need to hide / Show or instantiate objects that switch back and forth , Try not to use SetActiveRecursively or active, Instead, use the method of moving the object far out of the camera range and back to its original position ;
• Minimize the use of modular operations and division operations , such as a/5f, It must be written as a*0.2f.
• For variables or methods that are not often called or changed, it is recommended to use Coroutines & Yield;
• Try to use integer numbers , because iPhone The floating-point calculation ability of is very poor ;
• Do not use native GUI Method ;
• Do not instantiate （Instantiate） object , Build the object pool in advance , And use Translate“ Generate ” object ;

Model optimization

• Merge shaders with the same map , Merge... Using the same shader Mesh;
• There is only one character map and shader , If more than one is necessary, separate the model into multiple parts ;
• Don't use too much of the skeletal system ;
• When using multiple roles , Separate the animation separately ;
• Use layer distance to control the display distance of the model ;
• Shadow actually contains two aspects: darkness and shadow , It is recommended to bake out the shadow effect when using real-time shadow , Don't use lights to adjust the darkness .
• Use fewer pixel lights and those that use pixel lights Shader;
• If hard shadows can solve the problem, don't use soft shadows , And use low resolution shadows that do not affect the effect ;
• Real time shadows are performance intensive , Minimize the distance of shadows ;
• If allowed, use linear fog in large scenes , This makes distant objects or shadows imperceptible , Therefore, the performance can be improved by reducing the distance between the camera and the shadow ;
• Use smooth groups to minimize the number of faces of the model ;
• If there are no lights or objects moving in the project, do not use real-time lights ;
• Surface of the water 、 Mirror and other real-time reflection / The effect of refraction is placed separately Water In the layer , And according to the real-time reflection / Adjust the range of refraction ;
• Collision has little effect on efficiency , But collision is still recommended Box、Sphere Colliding body ;
• Try to consider the use of building material balls Substance;
• Try to reflect all real-time / refraction （ Like water 、 Mirror 、 The floor and so on ） All assembled into one face ;
• False reflexes / Refraction does not need to use too large a resolution , commonly 6464 Can , No more than 256256;
• Shaders that need to be changed , It is recommended to instantiate one , Instead of using public shaders ;
• Place objects that do not require rays or collision events in IgnoreRaycast Layers ;
• Place the water surface or similar effect on Water Layers
• Place the object of the transparent channel in TransparentFX Layers ;
• Develop good labels （Tags）、 level （Hieratchy） And layers （Layer） Methodical habits , Put different objects on different labels or layers , The effective combination of the three will be very convenient by name 、 Categories and attributes to find ;
• adopt Stats and Profile View the aspects or objects that have the greatest impact on efficiency , Or use the method of disabling some models to see where the problem is ;
• Use occlusion culling （Occlusion Culling） Deal with big scenes , A more primitive class LOD（ Multi level detail ） technology , And be able to “ Division ” A model as a whole .

Other

• If lights and pixel lights are not used in the scene , Don't use normal mapping , Because the normal effect is only when there is a light source （Direct Light/Point Light/Angle Light/Pixel Light） It works only when .