文章目录
- 环境
- 优化
- 示例
- Texture Format : Alpha 8 和 shaderlab : _TextureSampleAdd
- shaderlab : _UIMaskSoftnessX 和 _UIMaskSoftnessY
- shaderlab _Color
- shader
- ARM Mobile Studio - Graphics Analyzer
- 优化前
- 优化后
环境
Unity : 2020.3.37f1
Pipeline : BRP
优化
做性能优化都是慢慢都扣出来的
当然，优先处理 top 热点的
但是一些就算不是 top，但是像素面积有多，overdraw 也多不可小觑
当然，如果你能找到性能热点，那还是优先分析一下哪些 shader 占的 cycles 最高，和 A, L/S, T, 最高
还有 register 的数量尽可能的少
示例
比如： UI/Default 的 shader
下面优化的是一些小部分的内容，不是 top 热点，优化原理是一样的
优化选项：
- 浮点精度
- 避免类型转换
- 避免 laster z （一般是 alpha test 导致，或是 sv_depth 的深度操作）
- 删除无用的效果代码
- 删除不明显的效果代码
- 优化写法（这个一般需要了解一些 编译器 无法识别的 编译优化时，需要特殊处理）
Texture Format : Alpha 8 和 shaderlab : _TextureSampleAdd
如果确保项目中没有使用到 texture 的格式为 Alpha 8 的 UI 纹理，那么可以将 _TextureSampleAdd 删除
但是 _TextureSampleAdd 文档根本没有说明
cginc 或是 shader 文件也没有表述，这就很 Unity
因此之前压根不敢删除
谷歌 搜索了一下，找到：
What does _TextureSampleAdd do?
然后确定我们项目中没使用到 Alpha 8 格式的 UI 纹理，那么就可以直接删除了
看到了回答：
shaderlab : _UIMaskSoftnessX 和 _UIMaskSoftnessY
如果项目中的 RectMask2D 也没有使用到 软裁剪，也可以将 UI/Default 中的 _UIMaskSoftnessX 和 _UIMaskSoftnessY 删除
Softness 开启
Softness 关闭
shaderlab _Color
这块比较容易忽略，虽然只是 vs 的计算，而且 UI 的顶点数量不多，一般不是什么问题
但还是那句话，性能如果要做到极致，除了优化 top 热点，其余小的也是肉，能优化就优化
所以，如果项目中 UI/Default 使用了 _Color，但是项目中没去使用，那就可以删除
一般都是使用unity 自带的按钮状态来控制（这些是通过 vertex buffer 来控制的，因为这样可以 vertex buffer 合批）
shader
// Unity built-in shader source. Copyright (c) 2016 Unity Technologies. MIT license (see license.txt)
Shader "UI/Default"
{
Properties
{
[PerRendererData] _MainTex ("Sprite Texture", 2D) = "white" {}
_Color ("Tint", Color) = (1,1,1,1)
_StencilComp ("Stencil Comparison", Float) = 8
_Stencil ("Stencil ID", Float) = 0
_StencilOp ("Stencil Operation", Float) = 0
_StencilWriteMask ("Stencil Write Mask", Float) = 255
_StencilReadMask ("Stencil Read Mask", Float) = 255
_ColorMask ("Color Mask", Float) = 15
[Toggle(UNITY_UI_ALPHACLIP)] _UseUIAlphaClip ("Use Alpha Clip", Float) = 0
}
SubShader
{
Tags
{
"Queue"="Transparent"
"IgnoreProjector"="True"
"RenderType"="Transparent"
"PreviewType"="Plane"
"CanUseSpriteAtlas"="True"
}
Stencil
{
Ref [_Stencil]
Comp [_StencilComp]
Pass [_StencilOp]
ReadMask [_StencilReadMask]
WriteMask [_StencilWriteMask]
}
Cull Off
Lighting Off
ZWrite Off
ZTest [unity_GUIZTestMode]
Blend One OneMinusSrcAlpha
ColorMask [_ColorMask]
Pass
{
Name "Default"
CGPROGRAM
#pragma vertex vert
#pragma fragment frag
#pragma target 2.0
#include "UnityCG.cginc"
#include "UnityUI.cginc"
#pragma multi_compile_local _ UNITY_UI_CLIP_RECT
#pragma multi_compile_local _ UNITY_UI_ALPHACLIP
struct appdata_t
{
float4 vertex : POSITION;
float4 color : COLOR;
float2 texcoord : TEXCOORD0;
UNITY_VERTEX_INPUT_INSTANCE_ID
};
struct v2f
{
float4 vertex : SV_POSITION;
fixed4 color : COLOR;
float2 texcoord : TEXCOORD0;
float4 worldPosition : TEXCOORD1;
float4 mask : TEXCOORD2;
UNITY_VERTEX_OUTPUT_STEREO
};
sampler2D _MainTex;
fixed4 _Color;
fixed4 _TextureSampleAdd;
float4 _ClipRect;
float4 _MainTex_ST;
float _UIMaskSoftnessX;
float _UIMaskSoftnessY;
v2f vert(appdata_t v)
{
v2f OUT;
UNITY_SETUP_INSTANCE_ID(v);
UNITY_INITIALIZE_VERTEX_OUTPUT_STEREO(OUT);
float4 vPosition = UnityObjectToClipPos(v.vertex);
OUT.worldPosition = v.vertex;
OUT.vertex = vPosition;
float2 pixelSize = vPosition.w;
pixelSize /= float2(1, 1) * abs(mul((float2x2)UNITY_MATRIX_P, _ScreenParams.xy));
float4 clampedRect = clamp(_ClipRect, -2e10, 2e10);
float2 maskUV = (v.vertex.xy - clampedRect.xy) / (clampedRect.zw - clampedRect.xy);
OUT.texcoord = TRANSFORM_TEX(v.texcoord.xy, _MainTex);
OUT.mask = float4(v.vertex.xy * 2 - clampedRect.xy - clampedRect.zw, 0.25 / (0.25 * half2(_UIMaskSoftnessX, _UIMaskSoftnessY) + abs(pixelSize.xy)));
OUT.color = v.color * _Color;
return OUT;
}
fixed4 frag(v2f IN) : SV_Target
{
half4 color = IN.color * (tex2D(_MainTex, IN.texcoord) + _TextureSampleAdd);
#ifdef UNITY_UI_CLIP_RECT
half2 m = saturate((_ClipRect.zw - _ClipRect.xy - abs(IN.mask.xy)) * IN.mask.zw);
color.a *= m.x * m.y;
#endif
#ifdef UNITY_UI_ALPHACLIP
clip (color.a - 0.001);
#endif
color.rgb *= color.a;
return color;
}
ENDCG
}
}
}
优化后
// Unity built-in shader source. Copyright (c) 2016 Unity Technologies. MIT license (see license.txt)
Shader "UI/Default_EXT"
{
Properties
{
[PerRendererData] _MainTex ("Sprite Texture", 2D) = "white" {}
_StencilComp ("Stencil Comparison", Float) = 8
_Stencil ("Stencil ID", Float) = 0
_StencilOp ("Stencil Operation", Float) = 0
_StencilWriteMask ("Stencil Write Mask", Float) = 255
_StencilReadMask ("Stencil Read Mask", Float) = 255
_ColorMask ("Color Mask", Float) = 15
[Toggle(UNITY_UI_ALPHACLIP)] _UseUIAlphaClip ("Use Alpha Clip", Float) = 0
}
SubShader
{
Tags
{
"Queue"="Transparent"
"IgnoreProjector"="True"
"RenderType"="Transparent"
"PreviewType"="Plane"
"CanUseSpriteAtlas"="True"
}
Stencil
{
Ref [_Stencil]
Comp [_StencilComp]
Pass [_StencilOp]
ReadMask [_StencilReadMask]
WriteMask [_StencilWriteMask]
}
Cull Off
Lighting Off
ZWrite Off
ZTest [unity_GUIZTestMode]
Blend One OneMinusSrcAlpha
ColorMask [_ColorMask]
Pass
{
Name "Default"
CGPROGRAM
#pragma vertex vert
#pragma fragment frag
#pragma target 2.0
#include "UnityCG.cginc"
#include "UnityUI.cginc"
#pragma multi_compile_local _ UNITY_UI_CLIP_RECT
#pragma multi_compile_local _ UNITY_UI_ALPHACLIP
struct appdata_t
{
float4 vertex : POSITION;
fixed4 color : COLOR;
half2 texcoord : TEXCOORD0;
UNITY_VERTEX_INPUT_INSTANCE_ID
};
struct v2f
{
float4 vertex : SV_POSITION;
fixed4 color : COLOR;
half2 texcoord : TEXCOORD0;
//float4 worldPosition : TEXCOORD1;
half2 mask : TEXCOORD2;
UNITY_VERTEX_OUTPUT_STEREO
};
sampler2D _MainTex;
float4 _ClipRect;
half4 _MainTex_ST;
half _UIMaskSoftnessX;
half _UIMaskSoftnessY;
v2f vert(appdata_t v)
{
v2f OUT;
UNITY_SETUP_INSTANCE_ID(v);
UNITY_INITIALIZE_VERTEX_OUTPUT_STEREO(OUT);
float4 vPosition = UnityObjectToClipPos(v.vertex);
//OUT.worldPosition = v.vertex;
OUT.vertex = vPosition;
half2 pixelSize = vPosition.w;
pixelSize /= half2(1, 1) * abs(mul((half2x2)UNITY_MATRIX_P, _ScreenParams.xy));
float4 clampedRect = clamp(_ClipRect, -2e10, 2e10);
//float2 maskUV = (v.vertex.xy - clampedRect.xy) / (clampedRect.zw - clampedRect.xy);
OUT.texcoord = TRANSFORM_TEX(v.texcoord.xy, _MainTex);
OUT.mask = half2(v.vertex.xy * 2 - clampedRect.xy - clampedRect.zw);
OUT.color = v.color;
return OUT;
}
fixed4 frag(v2f IN) : SV_Target
{
half4 color = IN.color * tex2D(_MainTex, IN.texcoord);
//color.rgb *= half3(1.0, 0.1, 0.1);
#ifdef UNITY_UI_CLIP_RECT
half2 m = saturate(_ClipRect.zw - _ClipRect.xy - abs(IN.mask.xy));
color.a *= m.x * m.y;
#endif
#ifdef UNITY_UI_ALPHACLIP
clip (color.a - 0.001);
#endif
color.rgb *= color.a;
return color;
}
ENDCG
}
}
}
ARM Mobile Studio - Graphics Analyzer
vivo 手机 mali gpu 的来测试
AGA 分析一波
优化前
vs:
#version 320 es
#define HLSLCC_ENABLE_UNIFORM_BUFFERS 1
#if HLSLCC_ENABLE_UNIFORM_BUFFERS
#define UNITY_UNIFORM
#else
#define UNITY_UNIFORM uniform
#endif
#define UNITY_SUPPORTS_UNIFORM_LOCATION 1
#if UNITY_SUPPORTS_UNIFORM_LOCATION
#define UNITY_LOCATION(x) layout(location = x)
#define UNITY_BINDING(x) layout(binding = x, std140)
#else
#define UNITY_LOCATION(x)
#define UNITY_BINDING(x) layout(std140)
#endif
uniform vec4 _ScreenParams;
uniform vec4 hlslcc_mtx4x4unity_ObjectToWorld[4];
uniform vec4 hlslcc_mtx4x4glstate_matrix_projection[4];
uniform vec4 hlslcc_mtx4x4unity_MatrixVP[4];
uniform mediump vec4 _Color;
uniform vec4 _ClipRect;
uniform vec4 _MainTex_ST;
uniform float _UIMaskSoftnessX;
uniform float _UIMaskSoftnessY;
in highp vec4 in_POSITION0;
in highp vec4 in_COLOR0;
in highp vec2 in_TEXCOORD0;
out mediump vec4 vs_COLOR0;
out highp vec2 vs_TEXCOORD0;
out highp vec4 vs_TEXCOORD1;
out highp vec4 vs_TEXCOORD2;
vec4 u_xlat0;
vec4 u_xlat1;
void main()
{
u_xlat0 = in_POSITION0.yyyy * hlslcc_mtx4x4unity_ObjectToWorld[1];
u_xlat0 = hlslcc_mtx4x4unity_ObjectToWorld[0] * in_POSITION0.xxxx + u_xlat0;
u_xlat0 = hlslcc_mtx4x4unity_ObjectToWorld[2] * in_POSITION0.zzzz + u_xlat0;
u_xlat0 = u_xlat0 + hlslcc_mtx4x4unity_ObjectToWorld[3];
u_xlat1 = u_xlat0.yyyy * hlslcc_mtx4x4unity_MatrixVP[1];
u_xlat1 = hlslcc_mtx4x4unity_MatrixVP[0] * u_xlat0.xxxx + u_xlat1;
u_xlat1 = hlslcc_mtx4x4unity_MatrixVP[2] * u_xlat0.zzzz + u_xlat1;
u_xlat0 = hlslcc_mtx4x4unity_MatrixVP[3] * u_xlat0.wwww + u_xlat1;
gl_Position = u_xlat0;
u_xlat1 = in_COLOR0 * _Color;
vs_COLOR0 = u_xlat1;
vs_TEXCOORD0.xy = in_TEXCOORD0.xy * _MainTex_ST.xy + _MainTex_ST.zw;
vs_TEXCOORD1 = in_POSITION0;
u_xlat0.xy = _ScreenParams.yy * hlslcc_mtx4x4glstate_matrix_projection[1].xy;
u_xlat0.xy = hlslcc_mtx4x4glstate_matrix_projection[0].xy * _ScreenParams.xx + u_xlat0.xy;
u_xlat0.xy = u_xlat0.ww / abs(u_xlat0.xy);
u_xlat0.xy = vec2(_UIMaskSoftnessX, _UIMaskSoftnessY) * vec2(0.25, 0.25) + abs(u_xlat0.xy);
vs_TEXCOORD2.zw = vec2(0.25, 0.25) / u_xlat0.xy;
u_xlat0 = max(_ClipRect, vec4(-2e+10, -2e+10, -2e+10, -2e+10));
u_xlat0 = min(u_xlat0, vec4(2e+10, 2e+10, 2e+10, 2e+10));
u_xlat0.xy = in_POSITION0.xy * vec2(2.0, 2.0) + (-u_xlat0.xy);
vs_TEXCOORD2.xy = (-u_xlat0.zw) + u_xlat0.xy;
return;
}
FS:
#version 320 es
precision highp float;
precision highp int;
#define HLSLCC_ENABLE_UNIFORM_BUFFERS 1
#if HLSLCC_ENABLE_UNIFORM_BUFFERS
#define UNITY_UNIFORM
#else
#define UNITY_UNIFORM uniform
#endif
#define UNITY_SUPPORTS_UNIFORM_LOCATION 1
#if UNITY_SUPPORTS_UNIFORM_LOCATION
#define UNITY_LOCATION(x) layout(location = x)
#define UNITY_BINDING(x) layout(binding = x, std140)
#else
#define UNITY_LOCATION(x)
#define UNITY_BINDING(x) layout(std140)
#endif
uniform mediump vec4 _TextureSampleAdd;
uniform vec4 _ClipRect;
UNITY_LOCATION(0) uniform mediump sampler2D _MainTex;
in mediump vec4 vs_COLOR0;
in highp vec2 vs_TEXCOORD0;
in highp vec4 vs_TEXCOORD2;
layout(location = 0) out mediump vec4 SV_Target0;
vec4 u_xlat0;
mediump vec4 u_xlat16_0;
mediump float u_xlat16_1;
void main()
{
u_xlat0.xy = (-_ClipRect.xy) + _ClipRect.zw;
u_xlat0.xy = u_xlat0.xy + -abs(vs_TEXCOORD2.xy);
u_xlat0.xy = u_xlat0.xy * vs_TEXCOORD2.zw;
#ifdef UNITY_ADRENO_ES3
u_xlat0.xy = min(max(u_xlat0.xy, 0.0), 1.0);
#else
u_xlat0.xy = clamp(u_xlat0.xy, 0.0, 1.0);
#endif
u_xlat16_1 = u_xlat0.y * u_xlat0.x;
u_xlat16_0 = texture(_MainTex, vs_TEXCOORD0.xy);
u_xlat0 = u_xlat16_0 + _TextureSampleAdd;
u_xlat0 = u_xlat0 * vs_COLOR0;
u_xlat16_1 = u_xlat16_1 * u_xlat0.w;
SV_Target0.xyz = u_xlat0.xyz * vec3(u_xlat16_1);
SV_Target0.w = u_xlat16_1;
return;
}
优化后
VS：
Working Reigster 反而增加1就有点迷
#version 320 es
#define HLSLCC_ENABLE_UNIFORM_BUFFERS 1
#if HLSLCC_ENABLE_UNIFORM_BUFFERS
#define UNITY_UNIFORM
#else
#define UNITY_UNIFORM uniform
#endif
#define UNITY_SUPPORTS_UNIFORM_LOCATION 1
#if UNITY_SUPPORTS_UNIFORM_LOCATION
#define UNITY_LOCATION(x) layout(location = x)
#define UNITY_BINDING(x) layout(binding = x, std140)
#else
#define UNITY_LOCATION(x)
#define UNITY_BINDING(x) layout(std140)
#endif
uniform vec4 hlslcc_mtx4x4unity_ObjectToWorld[4];
uniform vec4 hlslcc_mtx4x4unity_MatrixVP[4];
uniform vec4 _ClipRect;
uniform mediump vec4 _MainTex_ST;
in highp vec4 in_POSITION0;
in mediump vec4 in_COLOR0;
in mediump vec2 in_TEXCOORD0;
out mediump vec4 vs_COLOR0;
out mediump vec2 vs_TEXCOORD0;
out mediump vec2 vs_TEXCOORD2;
vec4 u_xlat0;
vec4 u_xlat1;
void main()
{
u_xlat0 = in_POSITION0.yyyy * hlslcc_mtx4x4unity_ObjectToWorld[1];
u_xlat0 = hlslcc_mtx4x4unity_ObjectToWorld[0] * in_POSITION0.xxxx + u_xlat0;
u_xlat0 = hlslcc_mtx4x4unity_ObjectToWorld[2] * in_POSITION0.zzzz + u_xlat0;
u_xlat0 = u_xlat0 + hlslcc_mtx4x4unity_ObjectToWorld[3];
u_xlat1 = u_xlat0.yyyy * hlslcc_mtx4x4unity_MatrixVP[1];
u_xlat1 = hlslcc_mtx4x4unity_MatrixVP[0] * u_xlat0.xxxx + u_xlat1;
u_xlat1 = hlslcc_mtx4x4unity_MatrixVP[2] * u_xlat0.zzzz + u_xlat1;
gl_Position = hlslcc_mtx4x4unity_MatrixVP[3] * u_xlat0.wwww + u_xlat1;
vs_COLOR0 = in_COLOR0;
u_xlat0 = max(_ClipRect, vec4(-2e+10, -2e+10, -2e+10, -2e+10));
u_xlat0 = min(u_xlat0, vec4(2e+10, 2e+10, 2e+10, 2e+10));
u_xlat0.xy = in_POSITION0.xy * vec2(2.0, 2.0) + (-u_xlat0.xy);
u_xlat0.xy = (-u_xlat0.zw) + u_xlat0.xy;
vs_TEXCOORD2.xy = u_xlat0.xy;
vs_TEXCOORD0.xy = in_TEXCOORD0.xy * _MainTex_ST.xy + _MainTex_ST.zw;
return;
}
FS：
#version 320 es
precision highp float;
precision highp int;
#define HLSLCC_ENABLE_UNIFORM_BUFFERS 1
#if HLSLCC_ENABLE_UNIFORM_BUFFERS
#define UNITY_UNIFORM
#else
#define UNITY_UNIFORM uniform
#endif
#define UNITY_SUPPORTS_UNIFORM_LOCATION 1
#if UNITY_SUPPORTS_UNIFORM_LOCATION
#define UNITY_LOCATION(x) layout(location = x)
#define UNITY_BINDING(x) layout(binding = x, std140)
#else
#define UNITY_LOCATION(x)
#define UNITY_BINDING(x) layout(std140)
#endif
uniform vec4 _ClipRect;
UNITY_LOCATION(0) uniform mediump sampler2D _MainTex;
in mediump vec4 vs_COLOR0;
in mediump vec2 vs_TEXCOORD0;
in mediump vec2 vs_TEXCOORD2;
layout(location = 0) out mediump vec4 SV_Target0;
vec4 u_xlat0;
mediump vec4 u_xlat16_0;
mediump float u_xlat16_1;
void main()
{
u_xlat0.xy = (-_ClipRect.xy) + _ClipRect.zw;
u_xlat0.xy = u_xlat0.xy + -abs(vs_TEXCOORD2.xy);
#ifdef UNITY_ADRENO_ES3
u_xlat0.xy = min(max(u_xlat0.xy, 0.0), 1.0);
#else
u_xlat0.xy = clamp(u_xlat0.xy, 0.0, 1.0);
#endif
u_xlat16_1 = u_xlat0.y * u_xlat0.x;
u_xlat16_0 = texture(_MainTex, vs_TEXCOORD0.xy);
u_xlat0 = u_xlat16_0 * vs_COLOR0;
u_xlat16_1 = u_xlat16_1 * u_xlat0.w;
SV_Target0.xyz = u_xlat0.xyz * vec3(u_xlat16_1);
SV_Target0.w = u_xlat16_1;
return;
}
