// see README here: // github.com/ColinLeung-NiloCat/UnityURPUnlitScreenSpaceDecalShader Shader "Universal Render Pipeline/NiloCat Extension/Screen Space Decal/Unlit" { Properties { [Header(Basic)] [MainTexture]_MainTex("Texture", 2D) = "white" {} [MainColor][HDR]_Color("_Color (default = 1,1,1,1)", Color) = (1,1,1,1) [Header(Blending)] // https://docs.unity3d.com/ScriptReference/Rendering.BlendMode.html [Enum(UnityEngine.Rendering.BlendMode)]_DecalSrcBlend("_DecalSrcBlend (default = SrcAlpha)", Int) = 5 // 5 = SrcAlpha [Enum(UnityEngine.Rendering.BlendMode)]_DecalDstBlend("_DecalDstBlend (default = OneMinusSrcAlpha)", Int) = 10 // 10 = OneMinusSrcAlpha [Header(Alpha remap(extra alpha control))] _AlphaRemap("_AlphaRemap (default = 1,0,0,0) _____alpha will first mul x, then add y (zw unused)", vector) = (1,0,0,0) [Header(Prevent Side Stretching(Compare projection direction with scene normal and Discard if needed))] [Toggle(_ProjectionAngleDiscardEnable)] _ProjectionAngleDiscardEnable("_ProjectionAngleDiscardEnable (default = off)", float) = 0 _ProjectionAngleDiscardThreshold("_ProjectionAngleDiscardThreshold (default = 0)", range(-1,1)) = 0 [Header(Mul alpha to rgb)] [Toggle]_MulAlphaToRGB("_MulAlphaToRGB (default = off)", Float) = 0 [Header(Ignore texture wrap mode setting)] [Toggle(_FracUVEnable)] _FracUVEnable("_FracUVEnable (default = off)", Float) = 0 //====================================== below = usually can ignore in normal use case ===================================================================== [Header(Stencil Masking)] // https://docs.unity3d.com/ScriptReference/Rendering.CompareFunction.html _StencilRef("_StencilRef", Float) = 0 [Enum(UnityEngine.Rendering.CompareFunction)]_StencilComp("_StencilComp (default = Disable) _____Set to NotEqual if you want to mask by specific _StencilRef value, else set to Disable", Float) = 0 //0 = disable [Header(ZTest)] // https://docs.unity3d.com/ScriptReference/Rendering.CompareFunction.html // default need to be Disable, because we need to make sure decal render correctly even if camera goes into decal cube volume, although disable ZTest by default will prevent EarlyZ (bad for GPU performance) [Enum(UnityEngine.Rendering.CompareFunction)]_ZTest("_ZTest (default = Disable) _____to improve GPU performance, Set to LessEqual if camera never goes into cube volume, else set to Disable", Float) = 0 //0 = disable [Header(Cull)] // https://docs.unity3d.com/ScriptReference/Rendering.CullMode.html // default need to be Front, because we need to make sure decal render correctly even if camera goes into decal cube [Enum(UnityEngine.Rendering.CullMode)]_Cull("_Cull (default = Front) _____to improve GPU performance, Set to Back if camera never goes into cube volume, else set to Front", Float) = 1 //1 = Front [Header(Unity Fog)] [Toggle(_UnityFogEnable)] _UnityFogEnable("_UnityFogEnable (default = on)", Float) = 1 [Header(Support Orthographic camera)] [Toggle(_SupportOrthographicCamera)] _SupportOrthographicCamera("_SupportOrthographicCamera (default = off)", Float) = 0 } SubShader { // To avoid render order problems, Queue must >= 2501, which enters the transparent queue, // in transparent queue Unity will always draw from back to front // https://github.com/ColinLeung-NiloCat/UnityURPUnlitScreenSpaceDecalShader/issues/6#issuecomment-615940985 // https://docs.unity3d.com/Manual/SL-SubShaderTags.html // Queues up to 2500 (“Geometry+500”) are consided “opaque” and optimize the drawing order of the objects for best performance. // Higher rendering queues are considered for “transparent objects” and sort objects by distance, // starting rendering from the furthest ones and ending with the closest ones. // Skyboxes are drawn in between all opaque and all transparent objects. // "Queue" = "Transparent-499" means "Queue" = "2501", which is almost equals "draw right before any transparent objects" // "DisableBatching" means disable "dynamic batching", not "srp batching" Tags { "RenderType" = "Overlay" "Queue" = "Transparent-499" "DisableBatching" = "True" } Pass { Stencil { Ref[_StencilRef] Comp[_StencilComp] } Cull[_Cull] ZTest[_ZTest] ZWrite off Blend[_DecalSrcBlend][_DecalDstBlend] HLSLPROGRAM #pragma vertex vert #pragma fragment frag // make fog work #pragma multi_compile_fog // due to using ddx() & ddy() #pragma target 3.0 #pragma shader_feature_local_fragment _ProjectionAngleDiscardEnable #pragma shader_feature_local _UnityFogEnable #pragma shader_feature_local_fragment _FracUVEnable #pragma shader_feature_local_fragment _SupportOrthographicCamera // Required by all Universal Render Pipeline shaders. // It will include Unity built-in shader variables (except the lighting variables) // (https://docs.unity3d.com/Manual/SL-UnityShaderVariables.html // It will also include many utilitary functions. #include "Packages/com.unity.render-pipelines.universal/ShaderLibrary/Core.hlsl" struct appdata { float3 positionOS : POSITION; }; struct v2f { float4 positionCS : SV_POSITION; float4 screenPos : TEXCOORD0; float4 viewRayOS : TEXCOORD1; // xyz: viewRayOS, w: extra copy of positionVS.z float4 cameraPosOSAndFogFactor : TEXCOORD2; }; sampler2D _MainTex; sampler2D _CameraDepthTexture; CBUFFER_START(UnityPerMaterial) float4 _MainTex_ST; float _ProjectionAngleDiscardThreshold; half4 _Color; half2 _AlphaRemap; half _MulAlphaToRGB; CBUFFER_END v2f vert(appdata input) { v2f o; // VertexPositionInputs contains position in multiple spaces (world, view, homogeneous clip space, ndc) // Unity compiler will strip all unused references (say you don't use view space). // Therefore there is more flexibility at no additional cost with this struct. VertexPositionInputs vertexPositionInput = GetVertexPositionInputs(input.positionOS); o.positionCS = vertexPositionInput.positionCS; // regular unity fog #if _UnityFogEnable o.cameraPosOSAndFogFactor.a = ComputeFogFactor(o.positionCS.z); #else o.cameraPosOSAndFogFactor.a = 0; #endif // prepare depth texture's screen space UV o.screenPos = ComputeScreenPos(o.positionCS); // get "camera to vertex" ray in View space float3 viewRay = vertexPositionInput.positionVS; // [important note] //========================================================= // "viewRay z division" must do in the fragment shader, not vertex shader! (due to rasteriazation varying interpolation's perspective correction) // We skip the "viewRay z division" in vertex shader for now, and store the division value into varying o.viewRayOS.w first, // we will do the division later when we enter fragment shader // viewRay /= viewRay.z; //skip the "viewRay z division" in vertex shader for now o.viewRayOS.w = viewRay.z;//store the division value to varying o.viewRayOS.w //========================================================= // unity's camera space is right hand coord(negativeZ pointing into screen), we want positive z ray in fragment shader, so negate it viewRay *= -1; // it is ok to write very expensive code in decal's vertex shader, // it is just a unity cube(4*6 vertices) per decal only, won't affect GPU performance at all. float4x4 ViewToObjectMatrix = mul(UNITY_MATRIX_I_M, UNITY_MATRIX_I_V); // transform everything to object space(decal space) in vertex shader first, so we can skip all matrix mul() in fragment shader o.viewRayOS.xyz = mul((float3x3)ViewToObjectMatrix, viewRay); o.cameraPosOSAndFogFactor.xyz = mul(ViewToObjectMatrix, float4(0,0,0,1)).xyz; // hard code 0 or 1 can enable many compiler optimization return o; } // copied from URP12.1.2's ShaderVariablesFunctions.hlsl #if SHADER_LIBRARY_VERSION_MAJOR < 12 float LinearDepthToEyeDepth(float rawDepth) { #if UNITY_REVERSED_Z return _ProjectionParams.z - (_ProjectionParams.z - _ProjectionParams.y) * rawDepth; #else return _ProjectionParams.y + (_ProjectionParams.z - _ProjectionParams.y) * rawDepth; #endif } #endif half4 frag(v2f i) : SV_Target { // [important note] //======================================================================== // now do "viewRay z division" that we skipped in vertex shader earlier. i.viewRayOS.xyz /= i.viewRayOS.w; //======================================================================== float2 screenSpaceUV = i.screenPos.xy / i.screenPos.w; float sceneRawDepth = tex2D(_CameraDepthTexture, screenSpaceUV).r; float3 decalSpaceScenePos; #if _SupportOrthographicCamera // we have to support both orthographic and perspective camera projection // static uniform branch depends on unity_OrthoParams.w // (should we use UNITY_BRANCH here?) decided NO because https://forum.unity.com/threads/correct-use-of-unity_branch.476804/ if(unity_OrthoParams.w) { float sceneDepthVS = LinearDepthToEyeDepth(sceneRawDepth); //***Used a few lines from Asset: Lux URP Essentials by forst*** // Edit: The copied Lux URP stopped working at some point, and no one even knew why it worked in the first place //---------------------------------------------------------------------------- float2 viewRayEndPosVS_xy = float2(unity_OrthoParams.xy * (i.screenPos.xy - 0.5) * 2 /* to clip space */); // Ortho near/far plane xy pos float4 vposOrtho = float4(viewRayEndPosVS_xy, -sceneDepthVS, 1); // Constructing a view space pos float3 wposOrtho = mul(UNITY_MATRIX_I_V, vposOrtho).xyz; // Trans. view space to world space //---------------------------------------------------------------------------- // transform world to object space(decal space) decalSpaceScenePos = mul(GetWorldToObjectMatrix(), float4(wposOrtho, 1)).xyz; } else { #endif // if perspective camera, LinearEyeDepth will handle everything for user // remember we can't use LinearEyeDepth for orthographic camera! float sceneDepthVS = LinearEyeDepth(sceneRawDepth,_ZBufferParams); // scene depth in any space = rayStartPos + rayDir * rayLength // here all data in ObjectSpace(OS) or DecalSpace // be careful, viewRayOS is not a unit vector, so don't normalize it, it is a direction vector which view space z's length is 1 decalSpaceScenePos = i.cameraPosOSAndFogFactor.xyz + i.viewRayOS.xyz * sceneDepthVS; #if _SupportOrthographicCamera } #endif // convert unity cube's [-0.5,0.5] vertex pos range to [0,1] uv. Only works if you use a unity cube in mesh filter! float2 decalSpaceUV = decalSpaceScenePos.xy + 0.5; // discard logic //=================================================== // discard "out of cube volume" pixels float shouldClip = 0; #if _ProjectionAngleDiscardEnable // also discard "scene normal not facing decal projector direction" pixels float3 decalSpaceHardNormal = normalize(cross(ddx(decalSpaceScenePos), ddy(decalSpaceScenePos)));//reconstruct scene hard normal using scene pos ddx&ddy // compare scene hard normal with decal projector's dir, decalSpaceHardNormal.z equals dot(decalForwardDir,sceneHardNormalDir) shouldClip = decalSpaceHardNormal.z > _ProjectionAngleDiscardThreshold ? 0 : 1; #endif // call discard // if ZWrite is Off, clip() is fast enough on mobile, because it won't write the DepthBuffer, so no GPU pipeline stall(confirmed by ARM staff). clip(0.5 - abs(decalSpaceScenePos) - shouldClip); //=================================================== // sample the decal texture float2 uv = decalSpaceUV.xy * _MainTex_ST.xy + _MainTex_ST.zw;//Texture tiling & offset #if _FracUVEnable uv = frac(uv);// add frac to ignore texture wrap setting #endif half4 col = tex2D(_MainTex, uv); col *= _Color;// tint color col.a = saturate(col.a * _AlphaRemap.x + _AlphaRemap.y);// alpha remap MAD col.rgb *= lerp(1, col.a, _MulAlphaToRGB);// extra multiply alpha to RGB #if _UnityFogEnable // Mix the pixel color with fogColor. You can optionaly use MixFogColor to override the fogColor // with a custom one. col.rgb = MixFog(col.rgb, i.cameraPosOSAndFogFactor.a); #endif return col; } ENDHLSL } } }