Merge Hd/bugfix #4830

Author: sebastienlagarde

com.unity.render-pipelines.high-definition/CHANGELOG.md

@@ -86,6 +86,10 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 - Fixed CustomPassUtils scaling issues when used with RTHandles allocated from a RenderTexture.
 - Fixed undo of some properties on light editor.
 - Fixed material Emission properties not begin animated when recording an animation (case 1328108).
+- Fixed volumetric fog in planar reflections.
+- Fixed error with motion blur and small render targets.
+- Fixed issue with on-demand directional shadow maps looking broken when a reflection probe is updated at the same time.
+- Fixed cropping issue with the compositor camera bridge (case 1340549).
 
 ### Changed
 - Reduced the maximal number of bounces for both RTGI and RTR (case 1318876).
@@ -95,6 +99,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 - Changed ray tracing acceleration structure build, so that only meshes with HDRP materials are included (case 1322365).
 - Default black texture XR is now opaque (alpha = 1).
 - Changed default sidedness to double, when a mesh with a mix of single and double-sided materials is added to the ray tracing acceleration structure (case 1323451).
+- Updated the recursive rendering documentation (case 1338639).
 
 ## [10.4.0] - 2021-03-11
 

com.unity.render-pipelines.high-definition/Documentation~/Ray-Tracing-Recursive-Rendering.md

@@ -4,6 +4,8 @@ This feature is a replacement pipeline for rendering Meshes in the High Definiti
 
 The smoothness of a Material does not affect the way a ray reflects or refracts, which makes this rendering mode useful for rendering multi-layered transparent GameObjects.
 
+HDRP might display the sky color instead of a GameObject that has ray tracing applied. This happens when the GameObject is further away from the Camera than the Max Ray Length value set in the volume component. To make the GameObject appear correctly, increase the value of the Max Ray Length property.
+
 ![](Images/RayTracingRecursiveRendering1.png)
 
 **Car gear shift rendered with recursive ray tracing**
@@ -37,5 +39,5 @@ Since recursive rendering uses an independent render pass, HDRP cannot render an
 | -------------- | ------------------------------------------------------------ |
 | **LayerMask**  | Defines the layers that HDRP processes this ray-traced effect for. |
 | **Max Depth**  | Controls the maximum number of times a ray can reflect or refract before it stops and returns the final color. Increasing this value increases execution time exponentially. |
-| **Ray Length** | Controls the length of the rays that HDRP uses for ray tracing. If a ray doesn't find an intersection, then the ray returns the color of the sky. |
+| **Max Ray Length** | Controls the length of the rays that HDRP uses for ray tracing. If a ray doesn't find an intersection, then the ray returns the color of the sky. |
 | **Min Smoothness** | Defines the threshold at which reflection rays are not cast if the smoothness value of the target surface is inferior to the one defined by the parameter. |

com.unity.render-pipelines.high-definition/Documentation~/Shadows-in-HDRP.md

@@ -146,8 +146,8 @@ While you are in the Unity Editor, HDRP updates shadow maps whenever you modify
 - SetShadowResolutionOverride()
 - SetShadowUpdateMode() or shadowUpdateMode. In this case, HDRP only refreshes the cached shadow maps if the mode changes between Every Frame and not Every Frame).
 
-
 Be aware that anything that is view-dependent is likely to create problems with cached shadow maps because HDRP does not automatically update them as the main view moves. A non-obvious example of this is tessellation. Because tessellation factor is view-dependent, the geometry that the main camera sees might mismatch the geometry that HDRP rendered into the cached shadow map. If this visibly occurs, trigger a request for HDRP to update the Light's shadow map. To do this, make sure the Light's **Update Mode** is set to **On Demand** and call `RequestShadowMapRendering`.
+Another non obvious scenario is when multiple views are available, the light will be updated only for a single view therefore causing the other views to have incorrect results. To avoid a common scenario in which the described artifact will occur, HDRP will not mark a shadow request as completed when performed from reflection probes with view dependent shadows and waiting until a non-reflection camera triggers a shadow update.
 
 
 

com.unity.render-pipelines.high-definition/Runtime/Compositor/CompositionManager.cs

@@ -152,6 +152,8 @@ public bool shaderPropertiesAreDirty
 
         ShaderVariablesGlobal m_ShaderVariablesGlobalCB = new ShaderVariablesGlobal();
 
+        int m_RecorderTempRT = Shader.PropertyToID("TempRecorder");
+
         static private CompositionManager s_CompositorInstance;
 
         // Built-in Color.black has an alpha of 1, so defien here a fully transparent black 
@@ -824,14 +826,14 @@ void CustomRender(ScriptableRenderContext context, HDCamera camera)
                 if (recorderCaptureActions != null)
                 {
                     m_ShaderVariablesGlobalCB._ViewProjMatrix = m_ViewProjMatrixFlipped;
-                    cmd.SetInvertCulling(true);
                     ConstantBuffer.PushGlobal(cmd, m_ShaderVariablesGlobalCB, HDShaderIDs._ShaderVariablesGlobal);
-                    cmd.Blit(null, BuiltinRenderTextureType.CameraTarget, m_Material, m_Material.FindPass("ForwardOnly"));
+                    var format = m_InputLayers[0].GetRenderTarget().format;
+                    cmd.GetTemporaryRT(m_RecorderTempRT, camera.camera.pixelWidth, camera.camera.pixelHeight, 0, FilterMode.Point, format);
+                    cmd.Blit(null, m_RecorderTempRT, m_Material, m_Material.FindPass("ForwardOnly"));
                     for (recorderCaptureActions.Reset(); recorderCaptureActions.MoveNext();)
                     {
-                        recorderCaptureActions.Current(BuiltinRenderTextureType.CameraTarget, cmd);
+                        recorderCaptureActions.Current(m_RecorderTempRT, cmd);
                     }
-                    cmd.SetInvertCulling(false);
                 }
 
                 // When we render directly to game view, we render the image flipped up-side-down, like other HDRP cameras

com.unity.render-pipelines.high-definition/Runtime/Lighting/Light/HDAdditionalLightData.cs

@@ -2246,7 +2246,8 @@ internal int UpdateShadowRequest(HDCamera hdCamera, HDShadowManager manager, HDS
                 shadowRequest.cachedAtlasViewport = resolutionRequest.cachedAtlasViewport;
                 manager.UpdateShadowRequest(shadowRequestIndex, shadowRequest, updateType);
 
-                if (needToUpdateCachedContent)
+                if (needToUpdateCachedContent && (lightType != HDLightType.Directional ||
+                                                  hdCamera.camera.cameraType != CameraType.Reflection))
                 {
                     // Handshake with the cached shadow manager to notify about the rendering.
                     // Technically the rendering has not happened yet, but it is scheduled.

com.unity.render-pipelines.high-definition/Runtime/Lighting/VolumetricLighting/VolumetricLighting.cs

@@ -414,6 +414,10 @@ GenerateMaxZParameters PrepareGenerateMaxZParameters(HDCamera hdCamera, HDUtils.
         {
             var parameters = new GenerateMaxZParameters();
             parameters.generateMaxZCS = defaultResources.shaders.maxZCS;
+            parameters.generateMaxZCS.shaderKeywords = null;
+            bool planarReflection = hdCamera.camera.cameraType == CameraType.Reflection && hdCamera.parentCamera != null;
+            CoreUtils.SetKeyword(parameters.generateMaxZCS, "PLANAR_OBLIQUE_DEPTH", planarReflection);
+
             parameters.maxZKernel = parameters.generateMaxZCS.FindKernel("ComputeMaxZ");
             parameters.maxZDownsampleKernel = parameters.generateMaxZCS.FindKernel("ComputeFinalMask");
             parameters.dilateMaxZKernel = parameters.generateMaxZCS.FindKernel("DilateMask");

com.unity.render-pipelines.high-definition/Runtime/PostProcessing/PostProcessSystem.cs

@@ -2745,8 +2745,8 @@ MotionBlurParameters PrepareMotionBlurParameters(HDCamera camera)
                 tileSize = 16;
             }
 
-            int tileTexWidth = Mathf.CeilToInt(camera.actualWidth / tileSize);
-            int tileTexHeight = Mathf.CeilToInt(camera.actualHeight / tileSize);
+            int tileTexWidth = Mathf.CeilToInt(camera.actualWidth / (float)tileSize);
+            int tileTexHeight = Mathf.CeilToInt(camera.actualHeight / (float)tileSize);
             parameters.tileTargetSize = new Vector4(tileTexWidth, tileTexHeight, 1.0f / tileTexWidth, 1.0f / tileTexHeight);
 
             float screenMagnitude = (new Vector2(camera.actualWidth, camera.actualHeight).magnitude);

com.unity.render-pipelines.high-definition/Runtime/PostProcessing/Shaders/DepthOfFieldTileMax.compute

@@ -36,8 +36,13 @@ void KMain(uint2 dispatchThreadId : SV_DispatchThreadID, uint2 groupThreadId : S
 {
     uint threadIdx = groupThreadId.y * TILE_SIZE + groupThreadId.x;
 
+#if NEAR
     gs_cacheNear[threadIdx] = _InputNearCoCTexture[COORD_TEXTURE2D_X(min(dispatchThreadId, Size))].x;
+#endif
+    
+#if FAR    
     gs_cacheFar[threadIdx] = _InputFarCoCTexture[COORD_TEXTURE2D_X(min(dispatchThreadId, Size))].x;
+#endif
 
     GroupMemoryBarrierWithGroupSync();
 
@@ -47,8 +52,13 @@ void KMain(uint2 dispatchThreadId : SV_DispatchThreadID, uint2 groupThreadId : S
     {
         if (threadIdx < s)
         {
+#if NEAR        
             gs_cacheNear[threadIdx] = max(gs_cacheNear[threadIdx], gs_cacheNear[threadIdx + s]);
+#endif            
+
+#if FAR            
             gs_cacheFar[threadIdx] = max(gs_cacheFar[threadIdx], gs_cacheFar[threadIdx + s]);
+#endif            
         }
 
         GroupMemoryBarrierWithGroupSync();

com.unity.render-pipelines.high-definition/Runtime/RenderPipeline/RenderPass/GenerateMaxZ.compute

@@ -8,6 +8,7 @@
 #pragma kernel ComputeFinalMask FINAL_MASK=1
 #pragma kernel DilateMask DILATE_MASK=1
 
+#pragma multi_compile _ PLANAR_OBLIQUE_DEPTH
 
 // In some cases we might want to avoid stopping integrating volumetric even if > max distance if the gradient is very big.
 // Realistically, with the dilation step this was never seen as necessary. 
@@ -48,14 +49,20 @@ groupshared float gs_maxDepth[GROUP_SIZE * GROUP_SIZE];
 
 RW_TEXTURE2D_X(float, _OutputTexture);
 
+
 float GetDepthToDownsample(uint2 pixCoord)
 {
     float deviceDepth = LoadCameraDepth(pixCoord);
     float outputDepth = 0;
+
     if (deviceDepth == UNITY_RAW_FAR_CLIP_VALUE)
         outputDepth = 1e10f;
     else
+#ifdef PLANAR_OBLIQUE_DEPTH
+        outputDepth = ComputeViewSpacePosition(float2(pixCoord) * _ScreenSize.zw, deviceDepth, UNITY_MATRIX_I_P).z;
+#else
         outputDepth = LinearEyeDepth(LoadCameraDepth(pixCoord), _ZBufferParams);
+#endif
 
     return outputDepth;
 }