Unity3DGPUInstancing测试

GPU instancing 很早就支持手机了&＃xff08;Android只支持Opengl ES 3.0&＃xff09;&＃xff0c;最近在调研这个就对它测试了一下。

如果是不动的物体勾选static静态合并批次&＃xff08;40-50帧率&＃xff09;

自定义Shader中勾选Enable GPU Instancing

帧率竟然还不如静态合批次&＃xff08;帧率 30-40&＃xff09;

自定义Shader

Shader "SimplestInstancedShader1"

{

Properties

{

_Color ("Color", Color) &＃61; (1, 1, 1, 1)

_MainTex ("Texture", 2D) &＃61; "white" {}

}

SubShader

{

Tags { "RenderType"&＃61;"Opaque" }

LOD 100

Pass

{

CGPROGRAM

pragma vertex vert pragma fragment frag pragma multi_compile_instancing include "UnityCG.cginc"

struct appdata

{

float4 vertex : POSITION;

float2 uv : TEXCOORD0;

UNITY_VERTEX_INPUT_INSTANCE_ID

};

struct v2f

{

float2 uv : TEXCOORD0;

float4 vertex : SV_POSITION;

UNITY_VERTEX_INPUT_INSTANCE_ID // necessary only if you want to access instanced properties in fragment Shader.

};

UNITY_INSTANCING_BUFFER_START(Props)

UNITY_DEFINE_INSTANCED_PROP(float4, _Color)

UNITY_INSTANCING_BUFFER_END(Props)

sampler2D _MainTex;

float4 _MainTex_ST;

v2f vert(appdata v)

{

v2f o;

UNITY_SETUP_INSTANCE_ID(v);

UNITY_TRANSFER_INSTANCE_ID(v, o); // necessary only if you want to access instanced properties in the fragment Shader.

o.vertex &＃61; UnityObjectToClipPos(v.vertex);

o.uv &＃61; TRANSFORM_TEX(v.uv, _MainTex);

return o;

}

fixed4 frag(v2f i) : SV_Target

{

UNITY_SETUP_INSTANCE_ID(i); // necessary only if any instanced properties are going to be accessed in the fragment Shader.

return tex2D (_MainTex, i.uv) * UNITY_ACCESS_INSTANCED_PROP(Props, _Color);

}

ENDCG

}

}

}

所以&＃xff0c;静态不动的物体就没必要用上面的方法了&＃xff0c;于是我又测试了Graphics.DrawMeshInstanced()方法&＃xff0c;终于满意了。&＃xff08;稳定60帧&＃xff09;

Graphics.DrawMeshInstanced()方法不需要游戏对象以及游戏组件的额外开销&＃xff0c;在Update()方法中一气呵成&＃xff0c;不过它也有限制&＃xff0c;最多可以画1023个。

还有个方法是Graphics.DrawMeshInstancedIndirect()它没有画多少的限制&＃xff0c;而且更加灵活&＃xff0c;我也搞了好一会儿才在游戏中跑起来&＃xff0c;后来才知道它不能再手机上用&＃xff0c;只有PC上可以.

在shader中接收位置 颜色 矩阵

Shader "Instanced/InstancedIndirectSelection"

{

Properties

{

_MainTex ("Texture", 2D) &＃61; "white" {}

}

SubShader

{

Tags{ "RenderType" &＃61; "Opaque" }

LOD 100

Pass

{

CGPROGRAM

pragma vertex vert pragma fragment frag pragma target 4.5 include "UnityCG.cginc"

StructuredBuffer positionBuffer;

StructuredBuffer colorBuffer;

StructuredBuffer matrix4x4Buffer;

struct appdata

{

fixed4 color : COLOR;

float4 vertex : POSITION;

float4 texcoord : TEXCOORD0;

};

struct v2f

{

float4 color: COLOR;

float4 vertex : SV_POSITION;

float2 texcoord : TEXCOORD0;

};

sampler2D _MainTex;

float4 _MainTex_ST;

v2f vert(appdata v, uint instanceID : SV_InstanceID)

{

float4 data &＃61; positionBuffer[instanceID];

float4x4 materix &＃61; matrix4x4Buffer[instanceID / matrix4x4Buffer.Length];

float3 worldPosition &＃61; data.xyz &＃43; mul(materix,v.vertex.xyz * data.w);

v2f o;

o.vertex &＃61; mul(UNITY_MATRIX_VP, float4(worldPosition, 1.0f));

o.texcoord &＃61; TRANSFORM_TEX(v.texcoord, _MainTex);

o.color &＃61; colorBuffer[instanceID];

return o;

}

fixed4 frag(v2f i) : SV_Target

{

return tex2D (_MainTex, i.texcoord) * i.color;

}

ENDCG

}

}

}

在代码中将位置 颜色 矩阵传进去就行了

using UnityEngine;

public class InstancedIndirectExample : MonoBehaviour

{

public int instanceCount &＃61; 100000;

public Mesh instanceMesh;

public Material instanceMaterial;

private ComputeBuffer positionBuffer;

private ComputeBuffer argsBuffer;

private ComputeBuffer colorBuffer;

private ComputeBuffer matrix4x4Buffer;

private uint[] args &＃61; new uint[5] { 0, 0, 0, 0, 0 };

void Start()

{

argsBuffer &＃61; new ComputeBuffer(1, args.Length * sizeof(uint), ComputeBufferType.IndirectArguments);

UpdateBuffers();

}

void Update()

{

// instanceMaterial.SetBuffer("positionBuffer", positionBuffer);

Graphics.DrawMeshInstancedIndirect(instanceMesh, 0, instanceMaterial, new Bounds(Vector3.zero, new Vector3(100.0f, 100.0f, 100.0f)), argsBuffer);

}

void UpdateBuffers()

{

if ( instanceCount <1 ) instanceCount &＃61; 1;

// Positions & Colors

if (positionBuffer !&＃61; null) positionBuffer.Release();

if (colorBuffer !&＃61; null) colorBuffer.Release();

if (matrix4x4Buffer !&＃61; null) matrix4x4Buffer.Release();

positionBuffer &＃61; new ComputeBuffer(instanceCount, 16);

colorBuffer &＃61; new ComputeBuffer(instanceCount, 4*4);

matrix4x4Buffer &＃61; new ComputeBuffer(instanceCount,16);

Vector4[] positions &＃61; new Vector4[instanceCount];

Vector4[] colors &＃61; new Vector4[instanceCount];

Matrix4x4[] materix4X4 &＃61; new Matrix4x4[instanceCount];

for (int i&＃61;0; i

{

positions [i] &＃61; new Vector4 (i, 0f, 0f, 1f);

colors[i] &＃61; new Vector4( 1f, 1f, 1f, 1f );

materix4X4[i] &＃61; Matrix4x4.TRS (positions [i], Quaternion.Euler (0F, 0F, 0F), Vector3.one);

}

positionBuffer.SetData(positions);

colorBuffer.SetData(colors);

matrix4x4Buffer.SetData (materix4X4);

instanceMaterial.SetBuffer("positionBuffer", positionBuffer);

instanceMaterial.SetBuffer("colorBuffer", colorBuffer);

instanceMaterial.SetBuffer("matrix4x4Buffer", matrix4x4Buffer);

// indirect args

uint numIndices &＃61; (instanceMesh !&＃61; null) ? (uint)instanceMesh.GetIndexCount(0) : 0;

args[0] &＃61; numIndices;

args[1] &＃61; (uint)instanceCount;

argsBuffer.SetData(args);

}

void OnDisable()

{

if (positionBuffer !&＃61; null) positionBuffer.Release();

positionBuffer &＃61; null;

if (colorBuffer !&＃61; null) colorBuffer.Release();

colorBuffer &＃61; null;

if (argsBuffer !&＃61; null) argsBuffer.Release();

argsBuffer &＃61; null;

if (matrix4x4Buffer !&＃61; null) matrix4x4Buffer.Release();

matrix4x4Buffer &＃61; null;

}

}

如果是发生移动的物体顶点数量在900以内会动态合并批次&＃xff0c;如果需要支持更多的顶点GPU Instaning的优势就更明显了。 目前来看草、石头、植被、比较合适用它。