HDRP: Improve camera relative rendering precision on object

Improve camera relative rendering precision on object
- Apply camera translation on the model matrix
- Add a function GetObjectPositionWS
- User are required to always used TransformObjectToWorld or GetObjectToWorldMatrix,  they should never access to UNITY_MATRIX_M
- When a user want absolute world position it must use  GetAbsolutePositionWS(positionWS).  positionWS is either camera relative space or absolute position depends on settings, only a call to GetAbsolutePositionWS() guarantet that we are in absolute position

com.unity.render-pipelines.high-definition/HDRP/Material/Lit/LitDataMeshModification.hlsl

+// Note: positionWS can be either in camera relative space or not
 float3 GetVertexDisplacement(float3 positionWS, float3 normalWS, float2 texCoord0, float2 texCoord1, float2 texCoord2, float2 texCoord3, float4 vertexColor)
 {
     // This call will work for both LayeredLit and Lit shader
     return ComputePerVertexDisplacement(layerTexCoord, vertexColor, lod) * normalWS;
 }
 
+// Note: positionWS can be either in camera relative space or not
 void ApplyVertexModification(AttributesMesh input, float3 normalWS, inout float3 positionWS, float4 time)
 {
 #if defined(_VERTEX_DISPLACEMENT)
 #endif
 
 #ifdef _VERTEX_WIND
-    float3 rootWP = mul(GetObjectToWorldMatrix(), float4(0, 0, 0, 1)).xyz;
-    ApplyWindDisplacement(positionWS, normalWS, rootWP, _Stiffness, _Drag, _ShiverDrag, _ShiverDirectionality, _InitialBend, input.color.a, time);
+    // current wind implementation is in absolute world space
+    float3 rootWP = GetObjectPositionWS();
+    float3 absolutePositionWS = GetAbsolutePositionWS(positionWS);
+    ApplyWindDisplacement(absolutePositionWS, normalWS, rootWP, _Stiffness, _Drag, _ShiverDrag, _ShiverDirectionality, _InitialBend, input.color.a, time);
+    positionWS = GetCameraRelativePositionWS(absolutePositionWS);
 #endif
 }
 

com.unity.render-pipelines.high-definition/HDRP/ShaderPass/ShaderPassLightTransport.hlsl

     output.vmesh.positionCS = float4(uv * 2.0 - 1.0, inputMesh.positionOS.z > 0 ? 1.0e-4 : 0.0, 1.0);
 
 #ifdef VARYINGS_NEED_POSITION_WS
-    float3 positionWS = GetCameraRelativePositionWS(TransformObjectToWorld(inputMesh.positionOS));
+    float3 positionWS = TransformObjectToWorld(inputMesh.positionOS);
     output.vmesh.positionWS = positionWS;
 #endif
 

com.unity.render-pipelines.high-definition/HDRP/ShaderPass/VertMesh.hlsl

     UNITY_SETUP_INSTANCE_ID(input);
     UNITY_TRANSFER_INSTANCE_ID(input, output);
 
+    // This return the camera relative position (if enable)
     float3 positionWS = TransformObjectToWorld(input.positionOS);
 #ifdef ATTRIBUTES_NEED_NORMAL
     float3 normalWS = TransformObjectToWorldNormal(input.normalOS);
      float4 tangentWS = float4(TransformObjectToWorldDir(input.tangentOS.xyz), input.tangentOS.w);
 #endif
 
-    // TODO: deal with camera center rendering and instancing (This is the reason why we always perform two  steps transform to clip space + instancing matrix)
-
+     // Do vertex modification in camera relative space (if enable)
-
-    positionWS = GetCameraRelativePositionWS(positionWS);
 
 #ifdef TESSELLATION_ON
     output.positionWS = positionWS;

com.unity.render-pipelines.high-definition/HDRP/ShaderVariablesFunctions.hlsl

     return UNITY_MATRIX_V;
 }
 
+// Return world position of current object
+float3 GetObjectPositionWS()
+{
+    float4x4 modelMatrix = UNITY_MATRIX_M;
+    return modelMatrix._m03_m13_m23; // Translation object to world
+}
+
+// Return the PreTranslated ObjectToWorld Matrix (i.e matrix with _WorldSpaceCameraPos apply to it if we use camera relative rendering)
-    return UNITY_MATRIX_M;
+    float4x4 modelMatrix = UNITY_MATRIX_M;
+    // To handle camera relative rendering we substract the camera position in the model matrix
+    // User must not use UNITY_MATRIX_M directly, unless they understand what they do
+#if (SHADEROPTIONS_CAMERA_RELATIVE_RENDERING != 0)
+    modelMatrix._m03_m13_m23 -= _WorldSpaceCameraPos;
+#endif
+    return modelMatrix;
+#if (SHADEROPTIONS_CAMERA_RELATIVE_RENDERING != 0)
+    // To handle camera relative rendering we need to apply translation before converting to object space
+    float4x4 modelMatrix = { {1.0, 0.0, 0.0, _WorldSpaceCameraPos.x}, { 0.0, 1.0, 0.0, _WorldSpaceCameraPos.y }, { 0.0, 0.0, 1.0, _WorldSpaceCameraPos.z }, { 0.0, 0.0, 0.0, 1.0 } };
+    return mul(UNITY_MATRIX_I_M, modelMatrix);
+#else
+#endif
 }
 
 // Transform to homogenous clip space