Unity URP RenderTexture优化（二）：深度图优化

目录

前言：

一、定位深度信息

1.1：k_DepthStencilFormat

1.2：k_DepthBufferBits

1.3：_CameraDepthTexture精度与大小

1.4：_CameraDepthAttachment数量

二、全代码

---

前言：

在上一篇文章：Unity URP RenderTexture优化_unity urp优化-CSDN博客中的末尾我提到了优化_CameraDepthAttachment的建议，说实话其并不是最终的优化效果，当我们启用深度图开关后，依然会生成很多张深度副本，其大小依然时15MB左右，只有当关闭深度图渲染时，它们才会被修改的代码清除掉。若是项目需要深度信息，但这么多的RT资源对移动设备消耗太高，我们又该如何去优化它们呢？

好了，接下来我会以我的经验来优化它们。

本次文章核心优化UnityURP管线的深度信息，根据管线版本不同，也许URP的核心代码会有所不同，但我会讲解优化方向，以便于对于不同版本URP管线做到适配。以下版本为Unity2022.3.37f1.

一、定位深度信息

采用VSCode编码工具是个不错的选择，它不仅打开速度比VisualStudio快许多，查找内容也方便很多。我们依然需要修改UniversalRender脚本的源代码（UniversalRedner脚本路径为：Packages\com.unity.render-pipelines.universal@14.0.11\Runtime\UniversalRender【实际就是包文件路径的URP包里Runtime文件夹下】）

VSCode的Ctrl+F能帮助我快速找到含有depth的属性变量或者方法。

1.1：k_DepthStencilFormat

深度图的颜色格式，跟RenderTexture创建设置一样。

在URP源码中默认是 GraphicsFormat.D32_SFloat_S8_UInt，对于移动端来说我们并不需要这么高的颜色格式，可以修改为GraphicsFormat.D24_UNorm_S8_UInt或者是 D16_UNorm等D开头的GraphicsFormat，具体的需要查看格式是否适配，在GraphicsFormat中也有对各个格式的详细介绍：

在颜色格式精度低会导致深度信息不够，视觉效果就会有穿模现象，所以需要根据项目具体调整，这里我就根据配置按如下设置了

具体根据项目做调整

1.2：k_DepthBufferBits

URP默认是32位，此为深度Buffer位图精度，跟RenderTexture设置一样

数值需要设置为2的幂，常见为8，16，24，32。

同样如果太低可能会导致穿模问题，需要根据设置具体查看是否适配，我这里默认设置为16

1.3：_CameraDepthTexture精度与大小

这张图就表明了我对_CameraDepthTexture精度与大小的修改，红框中添加了一个对GraphicsFormat.R32_SFloat支持判断，因为有的设备可能不支持这么高的精度，不支持就可以走黄色部分的逻辑。

在红框中，我对颜色格式设置为无，精度都是按照我们上方设置好的，长宽都减半。

在黄框中，根据设备支持程度，颜色格式都有对应的适配调整，同样长度和宽度都减半。

这样你最后就会得到一个长宽减半，精度修改的_CameraDepthTexture。

1.4：_CameraDepthAttachment数量

上述实际并没有改变多少，大头仍然在深度图副本，它的数量太多，且大小和精度也太高。

核心在第一个红框的释放，能够大大减少副本的生成数量，第二个红框为正交相机深度的适配，看项目需求，若是没用到可以不修改，或者是在一开始就将k_DepthBufferBits改为24，因为16的精度太小了会导致穿模问题。

以上就是我对于深度图的优化，全代码如下（包含上篇文章的RT）

二、全代码

using System;
using System.Collections.Generic;
using System.Linq;
using UnityEngine.Experimental.Rendering;
using UnityEngine.Rendering.Universal.Internal;namespace UnityEngine.Rendering.Universal
{/// <summary>/// Rendering modes for Universal renderer./// </summary>public enum RenderingMode{/// <summary>Render all objects and lighting in one pass, with a hard limit on the number of lights that can be applied on an object.</summary>Forward = 0,/// <summary>Render all objects and lighting in one pass using a clustered data structure to access lighting data.</summary>[InspectorName("Forward+")]ForwardPlus = 2,/// <summary>Render all objects first in a g-buffer pass, then apply all lighting in a separate pass using deferred shading.</summary>Deferred = 1};/// <summary>/// When the Universal Renderer should use Depth Priming in Forward mode./// </summary>public enum DepthPrimingMode{/// <summary>Depth Priming will never be used.</summary>Disabled,/// <summary>Depth Priming will only be used if there is a depth prepass needed by any of the render passes.</summary>Auto,/// <summary>A depth prepass will be explicitly requested so Depth Priming can be used.</summary>Forced,}/// <summary>/// Default renderer for Universal RP./// This renderer is supported on all Universal RP supported platforms./// It uses a classic forward rendering strategy with per-object light culling./// </summary>public sealed partial class UniversalRenderer : ScriptableRenderer{
#if UNITY_IOS || UNITY_ANDROID || UNITY_EDITORconst GraphicsFormat k_DepthStencilFormat = GraphicsFormat.D24_UNorm_S8_UInt;const int k_DepthBufferBits = 16;
#elseconst GraphicsFormat k_DepthStencilFormat = GraphicsFormat.D32_SFloat_S8_UInt;const int k_DepthBufferBits = 32;
#endifconst int k_FinalBlitPassQueueOffset = 1;const int k_AfterFinalBlitPassQueueOffset = k_FinalBlitPassQueueOffset + 1;static readonly List<ShaderTagId> k_DepthNormalsOnly = new List<ShaderTagId> { new ShaderTagId("DepthNormalsOnly") };private static class Profiling{private const string k_Name = nameof(UniversalRenderer);public static readonly ProfilingSampler createCameraRenderTarget = new ProfilingSampler($"{k_Name}.{nameof(CreateCameraRenderTarget)}");}/// <inheritdoc/>public override int SupportedCameraStackingTypes(){switch (m_RenderingMode){case RenderingMode.Forward:case RenderingMode.ForwardPlus:return 1 << (int)CameraRenderType.Base | 1 << (int)CameraRenderType.Overlay;case RenderingMode.Deferred:return 1 << (int)CameraRenderType.Base;default:return 0;}}// Rendering mode setup from UI. The final rendering mode used can be different. See renderingModeActual.internal RenderingMode renderingModeRequested => m_RenderingMode;// Actual rendering mode, which may be different (ex: wireframe rendering, hardware not capable of deferred rendering).internal RenderingMode renderingModeActual => renderingModeRequested == RenderingMode.Deferred && (GL.wireframe || (DebugHandler != null && DebugHandler.IsActiveModeUnsupportedForDeferred) || m_DeferredLights == null || !m_DeferredLights.IsRuntimeSupportedThisFrame() || m_DeferredLights.IsOverlay)? RenderingMode.Forward: this.renderingModeRequested;bool m_Clustering;internal bool accurateGbufferNormals => m_DeferredLights != null ? m_DeferredLights.AccurateGbufferNormals : false;#if ADAPTIVE_PERFORMANCE_2_1_0_OR_NEWERinternal bool needTransparencyPass { get { return !UniversalRenderPipeline.asset.useAdaptivePerformance || !AdaptivePerformance.AdaptivePerformanceRenderSettings.SkipTransparentObjects;; } }
#endif/// <summary>Property to control the depth priming behavior of the forward rendering path.</summary>public DepthPrimingMode depthPrimingMode { get { return m_DepthPrimingMode; } set { m_DepthPrimingMode = value; } }DepthOnlyPass m_DepthPrepass;DepthNormalOnlyPass m_DepthNormalPrepass;CopyDepthPass m_PrimedDepthCopyPass;MotionVectorRenderPass m_MotionVectorPass;MainLightShadowCasterPass m_MainLightShadowCasterPass;AdditionalLightsShadowCasterPass m_AdditionalLightsShadowCasterPass;GBufferPass m_GBufferPass;CopyDepthPass m_GBufferCopyDepthPass;DeferredPass m_DeferredPass;DrawObjectsPass m_RenderOpaqueForwardOnlyPass;DrawObjectsPass m_RenderOpaqueForwardPass;DrawObjectsWithRenderingLayersPass m_RenderOpaqueForwardWithRenderingLayersPass;DrawSkyboxPass m_DrawSkyboxPass;CopyDepthPass m_CopyDepthPass;CopyColorPass m_CopyColorPass;TransparentSettingsPass m_TransparentSettingsPass;DrawObjectsPass m_RenderTransparentForwardPass;InvokeOnRenderObjectCallbackPass m_OnRenderObjectCallbackPass;FinalBlitPass m_FinalBlitPass;CapturePass m_CapturePass;
#if ENABLE_VR && ENABLE_XR_MODULEXROcclusionMeshPass m_XROcclusionMeshPass;CopyDepthPass m_XRCopyDepthPass;
#endif
#if UNITY_EDITORCopyDepthPass m_FinalDepthCopyPass;
#endifDrawScreenSpaceUIPass m_DrawOffscreenUIPass;DrawScreenSpaceUIPass m_DrawOverlayUIPass;internal RenderTargetBufferSystem m_ColorBufferSystem;internal RTHandle m_ActiveCameraColorAttachment;RTHandle m_ColorFrontBuffer;internal RTHandle m_ActiveCameraDepthAttachment;internal RTHandle m_CameraDepthAttachment;RTHandle m_XRTargetHandleAlias;internal RTHandle m_DepthTexture;RTHandle m_NormalsTexture;RTHandle m_DecalLayersTexture;RTHandle m_OpaqueColor;RTHandle m_MotionVectorColor;RTHandle m_MotionVectorDepth;ForwardLights m_ForwardLights;DeferredLights m_DeferredLights;RenderingMode m_RenderingMode;DepthPrimingMode m_DepthPrimingMode;CopyDepthMode m_CopyDepthMode;bool m_DepthPrimingRecommended;StencilState m_DefaultStencilState;LightCookieManager m_LightCookieManager;IntermediateTextureMode m_IntermediateTextureMode;bool m_VulkanEnablePreTransform;// Materials used in URP Scriptable Render PassesMaterial m_BlitMaterial = null;Material m_BlitHDRMaterial = null;Material m_CopyDepthMaterial = null;Material m_SamplingMaterial = null;Material m_StencilDeferredMaterial = null;Material m_CameraMotionVecMaterial = null;Material m_ObjectMotionVecMaterial = null;PostProcessPasses m_PostProcessPasses;internal ColorGradingLutPass colorGradingLutPass { get => m_PostProcessPasses.colorGradingLutPass; }internal PostProcessPass postProcessPass { get => m_PostProcessPasses.postProcessPass; }internal PostProcessPass finalPostProcessPass { get => m_PostProcessPasses.finalPostProcessPass; }internal RTHandle colorGradingLut { get => m_PostProcessPasses.colorGradingLut; }internal DeferredLights deferredLights { get => m_DeferredLights; }/// <summary>/// Constructor for the Universal Renderer./// </summary>/// <param name="data">The settings to create the renderer with.</param>public UniversalRenderer(UniversalRendererData data) : base(data){// Query and cache runtime platform info first before setting up URP.PlatformAutoDetect.Initialize();#if ENABLE_VR && ENABLE_XR_MODULEExperimental.Rendering.XRSystem.Initialize(XRPassUniversal.Create, data.xrSystemData.shaders.xrOcclusionMeshPS, data.xrSystemData.shaders.xrMirrorViewPS);
#endifm_BlitMaterial = CoreUtils.CreateEngineMaterial(data.shaders.coreBlitPS);m_BlitHDRMaterial = CoreUtils.CreateEngineMaterial(data.shaders.blitHDROverlay);m_CopyDepthMaterial = CoreUtils.CreateEngineMaterial(data.shaders.copyDepthPS);m_SamplingMaterial = CoreUtils.CreateEngineMaterial(data.shaders.samplingPS);m_StencilDeferredMaterial = CoreUtils.CreateEngineMaterial(data.shaders.stencilDeferredPS);m_CameraMotionVecMaterial = CoreUtils.CreateEngineMaterial(data.shaders.cameraMotionVector);m_ObjectMotionVecMaterial = CoreUtils.CreateEngineMaterial(data.shaders.objectMotionVector);StencilStateData stencilData = data.defaultStencilState;m_DefaultStencilState = StencilState.defaultValue;m_DefaultStencilState.enabled = stencilData.overrideStencilState;m_DefaultStencilState.SetCompareFunction(stencilData.stencilCompareFunction);m_DefaultStencilState.SetPassOperation(stencilData.passOperation);m_DefaultStencilState.SetFailOperation(stencilData.failOperation);m_DefaultStencilState.SetZFailOperation(stencilData.zFailOperation);m_IntermediateTextureMode = data.intermediateTextureMode;if (UniversalRenderPipeline.asset?.supportsLightCookies ?? false){var settings = LightCookieManager.Settings.Create();var asset = UniversalRenderPipeline.asset;if (asset){settings.atlas.format = asset.additionalLightsCookieFormat;settings.atlas.resolution = asset.additionalLightsCookieResolution;}m_LightCookieManager = new LightCookieManager(ref settings);}this.stripShadowsOffVariants = true;this.stripAdditionalLightOffVariants = true;
#if ENABLE_VR && ENABLE_VR_MODULE
#if PLATFORM_WINRT || PLATFORM_ANDROID// AdditionalLightOff variant is available on HL&Quest platform due to performance consideration.this.stripAdditionalLightOffVariants = !PlatformAutoDetect.isXRMobile;
#endif
#endifForwardLights.InitParams forwardInitParams;forwardInitParams.lightCookieManager = m_LightCookieManager;forwardInitParams.forwardPlus = data.renderingMode == RenderingMode.ForwardPlus;m_Clustering = data.renderingMode == RenderingMode.ForwardPlus;m_ForwardLights = new ForwardLights(forwardInitParams);//m_DeferredLights.LightCulling = data.lightCulling;this.m_RenderingMode = data.renderingMode;this.m_DepthPrimingMode = data.depthPrimingMode;this.m_CopyDepthMode = data.copyDepthMode;#if UNITY_ANDROID || UNITY_IOS || UNITY_TVOSthis.m_DepthPrimingRecommended = false;
#elsethis.m_DepthPrimingRecommended = true;
#endif// Note: Since all custom render passes inject first and we have stable sort,// we inject the builtin passes in the before events.m_MainLightShadowCasterPass = new MainLightShadowCasterPass(RenderPassEvent.BeforeRenderingShadows);m_AdditionalLightsShadowCasterPass = new AdditionalLightsShadowCasterPass(RenderPassEvent.BeforeRenderingShadows);#if ENABLE_VR && ENABLE_XR_MODULEm_XROcclusionMeshPass = new XROcclusionMeshPass(RenderPassEvent.BeforeRenderingOpaques);// Schedule XR copydepth right after m_FinalBlitPassm_XRCopyDepthPass = new CopyDepthPass(RenderPassEvent.AfterRendering + k_AfterFinalBlitPassQueueOffset, m_CopyDepthMaterial);
#endifm_DepthPrepass = new DepthOnlyPass(RenderPassEvent.BeforeRenderingPrePasses, RenderQueueRange.opaque, data.opaqueLayerMask);m_DepthNormalPrepass = new DepthNormalOnlyPass(RenderPassEvent.BeforeRenderingPrePasses, RenderQueueRange.opaque, data.opaqueLayerMask);if (renderingModeRequested == RenderingMode.Forward || renderingModeRequested == RenderingMode.ForwardPlus){m_PrimedDepthCopyPass = new CopyDepthPass(RenderPassEvent.AfterRenderingPrePasses, m_CopyDepthMaterial, true);}if (this.renderingModeRequested == RenderingMode.Deferred){var deferredInitParams = new DeferredLights.InitParams();deferredInitParams.stencilDeferredMaterial = m_StencilDeferredMaterial;deferredInitParams.lightCookieManager = m_LightCookieManager;m_DeferredLights = new DeferredLights(deferredInitParams, useRenderPassEnabled);m_DeferredLights.AccurateGbufferNormals = data.accurateGbufferNormals;m_GBufferPass = new GBufferPass(RenderPassEvent.BeforeRenderingGbuffer, RenderQueueRange.opaque, data.opaqueLayerMask, m_DefaultStencilState, stencilData.stencilReference, m_DeferredLights);// Forward-only pass only runs if deferred renderer is enabled.// It allows specific materials to be rendered in a forward-like pass.// We render both gbuffer pass and forward-only pass before the deferred lighting pass so we can minimize copies of depth buffer and// benefits from some depth rejection.// - If a material can be rendered either forward or deferred, then it should declare a UniversalForward and a UniversalGBuffer pass.// - If a material cannot be lit in deferred (unlit, bakedLit, special material such as hair, skin shader), then it should declare UniversalForwardOnly pass// - Legacy materials have unamed pass, which is implicitely renamed as SRPDefaultUnlit. In that case, they are considered forward-only too.// TO declare a material with unnamed pass and UniversalForward/UniversalForwardOnly pass is an ERROR, as the material will be rendered twice.StencilState forwardOnlyStencilState = DeferredLights.OverwriteStencil(m_DefaultStencilState, (int)StencilUsage.MaterialMask);ShaderTagId[] forwardOnlyShaderTagIds = new ShaderTagId[]{new ShaderTagId("UniversalForwardOnly"),new ShaderTagId("SRPDefaultUnlit"), // Legacy shaders (do not have a gbuffer pass) are considered forward-only for backward compatibilitynew ShaderTagId("LightweightForward") // Legacy shaders (do not have a gbuffer pass) are considered forward-only for backward compatibility};int forwardOnlyStencilRef = stencilData.stencilReference | (int)StencilUsage.MaterialUnlit;m_GBufferCopyDepthPass = new CopyDepthPass(RenderPassEvent.BeforeRenderingGbuffer + 1, m_CopyDepthMaterial, true);m_DeferredPass = new DeferredPass(RenderPassEvent.BeforeRenderingDeferredLights, m_DeferredLights);m_RenderOpaqueForwardOnlyPass = new DrawObjectsPass("Render Opaques Forward Only", forwardOnlyShaderTagIds, true, RenderPassEvent.BeforeRenderingOpaques, RenderQueueRange.opaque, data.opaqueLayerMask, forwardOnlyStencilState, forwardOnlyStencilRef);}// Always create this pass even in deferred because we use it for wireframe rendering in the Editor or offscreen depth texture rendering.m_RenderOpaqueForwardPass = new DrawObjectsPass(URPProfileId.DrawOpaqueObjects, true, RenderPassEvent.BeforeRenderingOpaques, RenderQueueRange.opaque, data.opaqueLayerMask, m_DefaultStencilState, stencilData.stencilReference);m_RenderOpaqueForwardWithRenderingLayersPass = new DrawObjectsWithRenderingLayersPass(URPProfileId.DrawOpaqueObjects, true, RenderPassEvent.BeforeRenderingOpaques, RenderQueueRange.opaque, data.opaqueLayerMask, m_DefaultStencilState, stencilData.stencilReference);bool copyDepthAfterTransparents = m_CopyDepthMode == CopyDepthMode.AfterTransparents;RenderPassEvent copyDepthEvent = copyDepthAfterTransparents ? RenderPassEvent.AfterRenderingTransparents : RenderPassEvent.AfterRenderingSkybox;m_CopyDepthPass = new CopyDepthPass(copyDepthEvent,m_CopyDepthMaterial,shouldClear: true,copyResolvedDepth: RenderingUtils.MultisampleDepthResolveSupported() && SystemInfo.supportsMultisampleAutoResolve && copyDepthAfterTransparents);// Motion vectors depend on the (copy) depth texture. Depth is reprojected to calculate motion vectors.m_MotionVectorPass = new MotionVectorRenderPass(copyDepthEvent + 1, m_CameraMotionVecMaterial, m_ObjectMotionVecMaterial, data.opaqueLayerMask);m_DrawSkyboxPass = new DrawSkyboxPass(RenderPassEvent.BeforeRenderingSkybox);m_CopyColorPass = new CopyColorPass(RenderPassEvent.AfterRenderingSkybox, m_SamplingMaterial, m_BlitMaterial);
#if ADAPTIVE_PERFORMANCE_2_1_0_OR_NEWERif (needTransparencyPass)
#endif{m_TransparentSettingsPass = new TransparentSettingsPass(RenderPassEvent.BeforeRenderingTransparents, data.shadowTransparentReceive);m_RenderTransparentForwardPass = new DrawObjectsPass(URPProfileId.DrawTransparentObjects, false, RenderPassEvent.BeforeRenderingTransparents, RenderQueueRange.transparent, data.transparentLayerMask, m_DefaultStencilState, stencilData.stencilReference);}m_OnRenderObjectCallbackPass = new InvokeOnRenderObjectCallbackPass(RenderPassEvent.BeforeRenderingPostProcessing);m_DrawOffscreenUIPass = new DrawScreenSpaceUIPass(RenderPassEvent.BeforeRenderingPostProcessing, true);m_DrawOverlayUIPass = new DrawScreenSpaceUIPass(RenderPassEvent.AfterRendering + k_AfterFinalBlitPassQueueOffset, false); // after m_FinalBlitPass{var postProcessParams = PostProcessParams.Create();postProcessParams.blitMaterial = m_BlitMaterial;postProcessParams.requestHDRFormat = GraphicsFormat.B10G11R11_UFloatPack32;var asset = UniversalRenderPipeline.asset;if (asset)postProcessParams.requestHDRFormat = UniversalRenderPipeline.MakeRenderTextureGraphicsFormat(asset.supportsHDR, asset.hdrColorBufferPrecision, false);m_PostProcessPasses = new PostProcessPasses(data.postProcessData, ref postProcessParams);}m_CapturePass = new CapturePass(RenderPassEvent.AfterRendering);m_FinalBlitPass = new FinalBlitPass(RenderPassEvent.AfterRendering + k_FinalBlitPassQueueOffset, m_BlitMaterial, m_BlitHDRMaterial);#if UNITY_EDITORm_FinalDepthCopyPass = new CopyDepthPass(RenderPassEvent.AfterRendering + 9, m_CopyDepthMaterial);
#endif// RenderTexture format depends on camera and pipeline (HDR, non HDR, etc)// Samples (MSAA) depend on camera and pipelinem_ColorBufferSystem = new RenderTargetBufferSystem("_CameraColorAttachment");supportedRenderingFeatures = new RenderingFeatures();if (this.renderingModeRequested == RenderingMode.Deferred){// Deferred rendering does not support MSAA.this.supportedRenderingFeatures.msaa = false;// Avoid legacy platforms: use vulkan instead.unsupportedGraphicsDeviceTypes = new GraphicsDeviceType[]{GraphicsDeviceType.OpenGLCore,GraphicsDeviceType.OpenGLES2,GraphicsDeviceType.OpenGLES3};}LensFlareCommonSRP.mergeNeeded = 0;LensFlareCommonSRP.maxLensFlareWithOcclusionTemporalSample = 1;LensFlareCommonSRP.Initialize();m_VulkanEnablePreTransform = GraphicsSettings.HasShaderDefine(BuiltinShaderDefine.UNITY_PRETRANSFORM_TO_DISPLAY_ORIENTATION);}/// <inheritdoc />protected override void Dispose(bool disposing){m_ForwardLights.Cleanup();m_GBufferPass?.Dispose();m_PostProcessPasses.Dispose();m_FinalBlitPass?.Dispose();m_DrawOffscreenUIPass?.Dispose();m_DrawOverlayUIPass?.Dispose();m_XRTargetHandleAlias?.Release();ReleaseRenderTargets();base.Dispose(disposing);CoreUtils.Destroy(m_BlitMaterial);CoreUtils.Destroy(m_BlitHDRMaterial);CoreUtils.Destroy(m_CopyDepthMaterial);CoreUtils.Destroy(m_SamplingMaterial);CoreUtils.Destroy(m_StencilDeferredMaterial);CoreUtils.Destroy(m_CameraMotionVecMaterial);CoreUtils.Destroy(m_ObjectMotionVecMaterial);CleanupRenderGraphResources();LensFlareCommonSRP.Dispose();}internal override void ReleaseRenderTargets(){m_ColorBufferSystem.Dispose();if (m_DeferredLights != null && !m_DeferredLights.UseRenderPass)m_GBufferPass?.Dispose();m_PostProcessPasses.ReleaseRenderTargets();m_MainLightShadowCasterPass?.Dispose();m_AdditionalLightsShadowCasterPass?.Dispose();m_CameraDepthAttachment?.Release();m_DepthTexture?.Release();m_NormalsTexture?.Release();m_DecalLayersTexture?.Release();m_OpaqueColor?.Release();m_MotionVectorColor?.Release();m_MotionVectorDepth?.Release();hasReleasedRTs = true;}private void SetupFinalPassDebug(ref CameraData cameraData){if ((DebugHandler != null) && DebugHandler.IsActiveForCamera(ref cameraData)){if (DebugHandler.TryGetFullscreenDebugMode(out DebugFullScreenMode fullScreenDebugMode, out int textureHeightPercent) &&(fullScreenDebugMode != DebugFullScreenMode.ReflectionProbeAtlas || m_Clustering)){Camera camera = cameraData.camera;float screenWidth = camera.pixelWidth;float screenHeight = camera.pixelHeight;var relativeSize = Mathf.Clamp01(textureHeightPercent / 100f);var height = relativeSize * screenHeight;var width = relativeSize * screenWidth;if (fullScreenDebugMode == DebugFullScreenMode.ReflectionProbeAtlas){// Ensure that atlas is not stretched, but doesn't take up more than the percentage in any dimension.var texture = m_ForwardLights.reflectionProbeManager.atlasRT;var targetWidth = height * texture.width / texture.height;if (targetWidth > width){height = width * texture.height / texture.width;}else{width = targetWidth;}}float normalizedSizeX = width / screenWidth;float normalizedSizeY = height / screenHeight;Rect normalizedRect = new Rect(1 - normalizedSizeX, 1 - normalizedSizeY, normalizedSizeX, normalizedSizeY);switch (fullScreenDebugMode){case DebugFullScreenMode.Depth:{DebugHandler.SetDebugRenderTarget(m_DepthTexture.nameID, normalizedRect, true);break;}case DebugFullScreenMode.AdditionalLightsShadowMap:{DebugHandler.SetDebugRenderTarget(m_AdditionalLightsShadowCasterPass.m_AdditionalLightsShadowmapHandle, normalizedRect, false);break;}case DebugFullScreenMode.MainLightShadowMap:{DebugHandler.SetDebugRenderTarget(m_MainLightShadowCasterPass.m_MainLightShadowmapTexture, normalizedRect, false);break;}case DebugFullScreenMode.ReflectionProbeAtlas:{DebugHandler.SetDebugRenderTarget(m_ForwardLights.reflectionProbeManager.atlasRT, normalizedRect, false);break;}default:{break;}}}else{DebugHandler.ResetDebugRenderTarget();}}}/// <summary>/// Returns if the camera renders to a offscreen depth texture./// </summary>/// <param name="cameraData">The camera data for the camera being rendered.</param>/// <returns>Returns true if the camera renders to depth without any color buffer. It will return false otherwise.</returns>public static bool IsOffscreenDepthTexture(in CameraData cameraData) => cameraData.targetTexture != null && cameraData.targetTexture.format == RenderTextureFormat.Depth;bool IsDepthPrimingEnabled(ref CameraData cameraData){// depth priming requires an extra depth copy, disable it on platforms not supporting it (like GLES when MSAA is on)if (!CanCopyDepth(ref cameraData))return false;// Depth Priming causes rendering errors with WebGL on Apple Arm64 GPUs.bool isNotWebGL = !IsWebGL();bool depthPrimingRequested = (m_DepthPrimingRecommended && m_DepthPrimingMode == DepthPrimingMode.Auto) || m_DepthPrimingMode == DepthPrimingMode.Forced;bool isForwardRenderingMode = m_RenderingMode == RenderingMode.Forward || m_RenderingMode == RenderingMode.ForwardPlus;bool isFirstCameraToWriteDepth = cameraData.renderType == CameraRenderType.Base || cameraData.clearDepth;// Enabled Depth priming when baking Reflection Probes causes artefacts (UUM-12397)bool isNotReflectionCamera = cameraData.cameraType != CameraType.Reflection;// Depth is not rendered in a depth-only camera setup with depth priming (UUM-38158)bool isNotOffscreenDepthTexture = !IsOffscreenDepthTexture(cameraData);return depthPrimingRequested && isForwardRenderingMode && isFirstCameraToWriteDepth && isNotReflectionCamera && isNotOffscreenDepthTexture && isNotWebGL;}bool IsWebGL(){
#if PLATFORM_WEBGLreturn IsGLESDevice();
#elsereturn false;
#endif}bool IsGLESDevice(){return SystemInfo.graphicsDeviceType == GraphicsDeviceType.OpenGLES2 || SystemInfo.graphicsDeviceType == GraphicsDeviceType.OpenGLES3;}bool IsGLDevice(){return IsGLESDevice() || SystemInfo.graphicsDeviceType == GraphicsDeviceType.OpenGLCore;}/// <inheritdoc />public override void Setup(ScriptableRenderContext context, ref RenderingData renderingData){m_ForwardLights.PreSetup(ref renderingData);ref CameraData cameraData = ref renderingData.cameraData;Camera camera = cameraData.camera;RenderTextureDescriptor cameraTargetDescriptor = cameraData.cameraTargetDescriptor;var cmd = renderingData.commandBuffer;if (DebugHandler != null){DebugHandler.Setup(context, ref renderingData);if (DebugHandler.IsActiveForCamera(ref cameraData)){if (DebugHandler.WriteToDebugScreenTexture(ref cameraData)){RenderTextureDescriptor colorDesc = cameraData.cameraTargetDescriptor;DebugHandler.ConfigureColorDescriptorForDebugScreen(ref colorDesc, cameraData.pixelWidth, cameraData.pixelHeight);RenderingUtils.ReAllocateIfNeeded(ref DebugHandler.DebugScreenColorHandle, colorDesc, name: "_DebugScreenColor");RenderTextureDescriptor depthDesc = cameraData.cameraTargetDescriptor;DebugHandler.ConfigureDepthDescriptorForDebugScreen(ref depthDesc, k_DepthBufferBits, cameraData.pixelWidth, cameraData.pixelHeight);RenderingUtils.ReAllocateIfNeeded(ref DebugHandler.DebugScreenDepthHandle, depthDesc, name: "_DebugScreenDepth");}if (DebugHandler.HDRDebugViewIsActive(ref cameraData)){DebugHandler.hdrDebugViewPass.Setup(ref cameraData, DebugHandler.DebugDisplaySettings.lightingSettings.hdrDebugMode);EnqueuePass(DebugHandler.hdrDebugViewPass);}}}if (cameraData.cameraType != CameraType.Game)useRenderPassEnabled = false;// Because of the shortcutting done by depth only offscreen cameras, useDepthPriming must be computed earlyuseDepthPriming = IsDepthPrimingEnabled(ref cameraData);// Special path for depth only offscreen cameras. Only write opaques + transparents.if (IsOffscreenDepthTexture(in cameraData)){ConfigureCameraTarget(k_CameraTarget, k_CameraTarget);SetupRenderPasses(in renderingData);EnqueuePass(m_RenderOpaqueForwardPass);#if ADAPTIVE_PERFORMANCE_2_1_0_OR_NEWERif (!needTransparencyPass)return;
#endifEnqueuePass(m_RenderTransparentForwardPass);return;}// Assign the camera color target early in case it is needed during AddRenderPasses.bool isPreviewCamera = cameraData.isPreviewCamera;var createColorTexture = ((rendererFeatures.Count != 0 && m_IntermediateTextureMode == IntermediateTextureMode.Always) && !isPreviewCamera) ||(Application.isEditor && m_Clustering);// Gather render passe input requirementsRenderPassInputSummary renderPassInputs = GetRenderPassInputs(ref renderingData);// Gather render pass require rendering layers event and mask sizebool requiresRenderingLayer = RenderingLayerUtils.RequireRenderingLayers(this, rendererFeatures,cameraTargetDescriptor.msaaSamples,out var renderingLayersEvent, out var renderingLayerMaskSize);// All passes that use write to rendering layers are excluded from gl// So we disable it to avoid setting multiple render targetsif (IsGLDevice())requiresRenderingLayer = false;bool renderingLayerProvidesByDepthNormalPass = false;bool renderingLayerProvidesRenderObjectPass = false;if (requiresRenderingLayer && renderingModeActual != RenderingMode.Deferred){switch (renderingLayersEvent){case RenderingLayerUtils.Event.DepthNormalPrePass:renderingLayerProvidesByDepthNormalPass = true;break;case RenderingLayerUtils.Event.Opaque:renderingLayerProvidesRenderObjectPass = true;break;default:throw new ArgumentOutOfRangeException();}}// Enable depth normal prepassif (renderingLayerProvidesByDepthNormalPass)renderPassInputs.requiresNormalsTexture = true;// TODO: investigate the order of call, had to change because of requiresRenderingLayerif (m_DeferredLights != null){m_DeferredLights.RenderingLayerMaskSize = renderingLayerMaskSize;m_DeferredLights.UseDecalLayers = requiresRenderingLayer;// TODO: This needs to be setup early, otherwise gbuffer attachments will be allocated with wrong sizem_DeferredLights.HasNormalPrepass = renderPassInputs.requiresNormalsTexture;m_DeferredLights.ResolveMixedLightingMode(ref renderingData);m_DeferredLights.IsOverlay = cameraData.renderType == CameraRenderType.Overlay;if (m_DeferredLights.UseRenderPass){// At this point we only have injected renderer features in the queue and can do assumptions on whether we'll need Framebuffer Fetchforeach (var pass in activeRenderPassQueue){if (pass.renderPassEvent >= RenderPassEvent.AfterRenderingGbuffer &&pass.renderPassEvent <= RenderPassEvent.BeforeRenderingDeferredLights){m_DeferredLights.DisableFramebufferFetchInput();break;}}}}// Should apply post-processing after rendering this camera?bool applyPostProcessing = cameraData.postProcessEnabled && m_PostProcessPasses.isCreated;// There's at least a camera in the camera stack that applies post-processingbool anyPostProcessing = renderingData.postProcessingEnabled && m_PostProcessPasses.isCreated;// If Camera's PostProcessing is enabled and if there any enabled PostProcessing requires depth texture as shader read resource (Motion Blur/DoF)bool cameraHasPostProcessingWithDepth = applyPostProcessing && cameraData.postProcessingRequiresDepthTexture;// TODO: We could cache and generate the LUT before rendering the stackbool generateColorGradingLUT = cameraData.postProcessEnabled && m_PostProcessPasses.isCreated;bool isSceneViewOrPreviewCamera = cameraData.isSceneViewCamera || cameraData.isPreviewCamera;// This indicates whether the renderer will output a depth texture.bool requiresDepthTexture = cameraData.requiresDepthTexture || renderPassInputs.requiresDepthTexture || m_DepthPrimingMode == DepthPrimingMode.Forced;#if UNITY_EDITORbool isGizmosEnabled = UnityEditor.Handles.ShouldRenderGizmos();
#elsebool isGizmosEnabled = false;
#endifbool mainLightShadows = m_MainLightShadowCasterPass.Setup(ref renderingData);bool additionalLightShadows = m_AdditionalLightsShadowCasterPass.Setup(ref renderingData);bool transparentsNeedSettingsPass = m_TransparentSettingsPass.Setup();bool forcePrepass = (m_CopyDepthMode == CopyDepthMode.ForcePrepass);// Depth prepass is generated in the following cases:// - If game or offscreen camera requires it we check if we can copy the depth from the rendering opaques pass and use that instead.// - Scene or preview cameras always require a depth texture. We do a depth pre-pass to simplify it and it shouldn't matter much for editor.// - Render passes require itbool requiresDepthPrepass = (requiresDepthTexture || cameraHasPostProcessingWithDepth) && (!CanCopyDepth(ref renderingData.cameraData) || forcePrepass);requiresDepthPrepass |= isSceneViewOrPreviewCamera;requiresDepthPrepass |= isGizmosEnabled;requiresDepthPrepass |= isPreviewCamera;requiresDepthPrepass |= renderPassInputs.requiresDepthPrepass;requiresDepthPrepass |= renderPassInputs.requiresNormalsTexture;// Current aim of depth prepass is to generate a copy of depth buffer, it is NOT to prime depth buffer and reduce overdraw on non-mobile platforms.// When deferred renderer is enabled, depth buffer is already accessible so depth prepass is not needed.// The only exception is for generating depth-normal textures: SSAO pass needs it and it must run before forward-only geometry.// DepthNormal prepass will render:// - forward-only geometry when deferred renderer is enabled// - all geometry when forward renderer is enabledif (requiresDepthPrepass && this.renderingModeActual == RenderingMode.Deferred && !renderPassInputs.requiresNormalsTexture)requiresDepthPrepass = false;requiresDepthPrepass |= useDepthPriming;// If possible try to merge the opaque and skybox passes instead of splitting them when "Depth Texture" is required.// The copying of depth should normally happen after rendering opaques.// But if we only require it for post processing or the scene camera then we do it after rendering transparent objects// Aim to have the most optimized render pass event for Depth Copy (The aim is to minimize the number of render passes)if (requiresDepthTexture){bool copyDepthAfterTransparents = m_CopyDepthMode == CopyDepthMode.AfterTransparents;RenderPassEvent copyDepthPassEvent = copyDepthAfterTransparents ? RenderPassEvent.AfterRenderingTransparents : RenderPassEvent.AfterRenderingOpaques;// RenderPassInputs's requiresDepthTexture is configured through ScriptableRenderPass's ConfigureInput functionif (renderPassInputs.requiresDepthTexture){// Do depth copy before the render pass that requires depth texture as shader read resourcecopyDepthPassEvent = (RenderPassEvent)Mathf.Min((int)RenderPassEvent.AfterRenderingTransparents, ((int)renderPassInputs.requiresDepthTextureEarliestEvent) - 1);}m_CopyDepthPass.renderPassEvent = copyDepthPassEvent;}else if (cameraHasPostProcessingWithDepth || isSceneViewOrPreviewCamera || isGizmosEnabled){// If only post process requires depth texture, we can re-use depth buffer from main geometry pass instead of enqueuing a depth copy pass, but no proper API to do that for now, so resort to depth copy pass for nowm_CopyDepthPass.renderPassEvent = RenderPassEvent.AfterRenderingTransparents;}createColorTexture |= RequiresIntermediateColorTexture(ref cameraData);createColorTexture |= renderPassInputs.requiresColorTexture;createColorTexture |= renderPassInputs.requiresColorTextureCreated;createColorTexture &= !isPreviewCamera;// If camera requires depth and there's no depth pre-pass we create a depth texture that can be read later by effect requiring it.// When deferred renderer is enabled, we must always create a depth texture and CANNOT use BuiltinRenderTextureType.CameraTarget. This is to get// around a bug where during gbuffer pass (MRT pass), the camera depth attachment is correctly bound, but during// deferred pass ("camera color" + "camera depth"), the implicit depth surface of "camera color" is used instead of "camera depth",// because BuiltinRenderTextureType.CameraTarget for depth means there is no explicit depth attachment...bool createDepthTexture = (requiresDepthTexture || cameraHasPostProcessingWithDepth) && !requiresDepthPrepass;createDepthTexture |= !cameraData.resolveFinalTarget;// Deferred renderer always need to access depth buffer.createDepthTexture |= (this.renderingModeActual == RenderingMode.Deferred && !useRenderPassEnabled);// Some render cases (e.g. Material previews) have shown we need to create a depth texture when we're forcing a prepass.createDepthTexture |= useDepthPriming;// Todo seems like with mrt depth is not taken from first targetcreateDepthTexture |= (renderingLayerProvidesRenderObjectPass);#if ENABLE_VR && ENABLE_XR_MODULE// URP can't handle msaa/size mismatch between depth RT and color RT(for now we create intermediate textures to ensure they match)if (cameraData.xr.enabled)createColorTexture |= createDepthTexture;
#endif
#if UNITY_ANDROID || UNITY_WEBGL// GLES can not use render texture's depth buffer with the color buffer of the backbuffer// in such case we create a color texture for it too.// If Vulkan PreTransform is enabled we can't mix backbuffer and intermediate render target due to screen orientation mismatchif (SystemInfo.graphicsDeviceType != GraphicsDeviceType.Vulkan || m_VulkanEnablePreTransform)createColorTexture |= createDepthTexture;
#endif// If there is any scaling, the color and depth need to be the same resolution and the target texture// will not be the proper size in this case. Same happens with GameView.// This introduces the final blit pass.if (RTHandles.rtHandleProperties.rtHandleScale.x != 1.0f || RTHandles.rtHandleProperties.rtHandleScale.y != 1.0f)createColorTexture |= createDepthTexture;if (useRenderPassEnabled || useDepthPriming)createColorTexture |= createDepthTexture;//Set rt descriptors so preview camera's have access should it be neededvar colorDescriptor = cameraTargetDescriptor;colorDescriptor.useMipMap = false;colorDescriptor.autoGenerateMips = false;colorDescriptor.depthBufferBits = (int)DepthBits.None;m_ColorBufferSystem.SetCameraSettings(colorDescriptor, FilterMode.Bilinear);bool clearInPostProcess = false;// Configure all settings require to start a new camera stack (base camera only)if (cameraData.renderType == CameraRenderType.Base){// Scene filtering redraws the objects on top of the resulting frame. It has to draw directly to the sceneview buffer.bool sceneViewFilterEnabled = camera.sceneViewFilterMode == Camera.SceneViewFilterMode.ShowFiltered;bool intermediateRenderTexture = (createColorTexture || createDepthTexture) && !sceneViewFilterEnabled;// RTHandles do not support combining color and depth in the same texture so we create them separately// Should be independent from filtered scene view//清除开关逻辑if (cameraData.targetTexture != null)createDepthTexture |= createColorTexture;else{if (VolumeManager.instance.GetVolumes(renderingData.cameraData.volumeLayerMask).Length <= 0)clearInPostProcess = true;else{Volume currentVolume = VolumeManager.instance.GetVolumes(renderingData.cameraData.volumeLayerMask)[0];if (currentVolume == null || !currentVolume.isActiveAndEnabled || !currentVolume.isGlobal || !currentVolume.gameObject.activeInHierarchy)clearInPostProcess = true;else{var stack = VolumeManager.instance.stack;var components = VolumeManager.instance.baseComponentTypeArray.Where(t => t.GetInterface(nameof(IPostProcessComponent)) != null && stack.GetComponent(t) != null).Select(t => stack.GetComponent(t) as IPostProcessComponent).ToList();if (components == null || components.Count == 0)clearInPostProcess = true;else{var active_components = new List<IPostProcessComponent>();foreach (var item in components){if (item.IsActive())active_components.Add(item);}if (active_components.Count == 0)clearInPostProcess = true;else{if (renderingData.cameraData.requiresDepthTexture || renderingData.cameraData.postProcessingRequiresDepthTexture)createDepthTexture |= createColorTexture;}}}}}RenderTargetIdentifier targetId = BuiltinRenderTextureType.CameraTarget;
#if ENABLE_VR && ENABLE_XR_MODULEif (cameraData.xr.enabled)targetId = cameraData.xr.renderTarget;
#endifif (m_XRTargetHandleAlias == null){m_XRTargetHandleAlias = RTHandles.Alloc(targetId);}else if (m_XRTargetHandleAlias.nameID != targetId){RTHandleStaticHelpers.SetRTHandleUserManagedWrapper(ref m_XRTargetHandleAlias, targetId);}// Doesn't create texture for Overlay cameras as they are already overlaying on top of created textures.if (intermediateRenderTexture)CreateCameraRenderTarget(context, ref cameraTargetDescriptor, useDepthPriming, cmd, ref cameraData);m_RenderOpaqueForwardPass.m_IsActiveTargetBackBuffer = !intermediateRenderTexture;m_RenderTransparentForwardPass.m_IsActiveTargetBackBuffer = !intermediateRenderTexture;m_DrawSkyboxPass.m_IsActiveTargetBackBuffer = !intermediateRenderTexture;
#if ENABLE_VR && ENABLE_XR_MODULEm_XROcclusionMeshPass.m_IsActiveTargetBackBuffer = !intermediateRenderTexture;
#endifm_ActiveCameraColorAttachment = createColorTexture ? m_ColorBufferSystem.PeekBackBuffer() : m_XRTargetHandleAlias;m_ActiveCameraDepthAttachment = createDepthTexture ? m_CameraDepthAttachment : m_XRTargetHandleAlias;}else{cameraData.baseCamera.TryGetComponent<UniversalAdditionalCameraData>(out var baseCameraData);var baseRenderer = (UniversalRenderer)baseCameraData.scriptableRenderer;if (m_ColorBufferSystem != baseRenderer.m_ColorBufferSystem){m_ColorBufferSystem.Dispose();m_ColorBufferSystem = baseRenderer.m_ColorBufferSystem;}m_ActiveCameraColorAttachment = m_ColorBufferSystem.PeekBackBuffer();m_ActiveCameraDepthAttachment = baseRenderer.m_ActiveCameraDepthAttachment;m_XRTargetHandleAlias = baseRenderer.m_XRTargetHandleAlias;}if (rendererFeatures.Count != 0 && !isPreviewCamera)ConfigureCameraColorTarget(m_ColorBufferSystem.PeekBackBuffer());bool copyColorPass = renderingData.cameraData.requiresOpaqueTexture || renderPassInputs.requiresColorTexture;// Check the createColorTexture logic above: intermediate color texture is not available for preview cameras.// Because intermediate color is not available and copyColor pass requires it, we disable CopyColor pass here.copyColorPass &= !isPreviewCamera;// Assign camera targets (color and depth)ConfigureCameraTarget(m_ActiveCameraColorAttachment, m_ActiveCameraDepthAttachment);bool hasPassesAfterPostProcessing = activeRenderPassQueue.Find(x => x.renderPassEvent == RenderPassEvent.AfterRenderingPostProcessing) != null;if (mainLightShadows)EnqueuePass(m_MainLightShadowCasterPass);if (additionalLightShadows)EnqueuePass(m_AdditionalLightsShadowCasterPass);bool requiresDepthCopyPass = !requiresDepthPrepass&& (renderingData.cameraData.requiresDepthTexture || cameraHasPostProcessingWithDepth || renderPassInputs.requiresDepthTexture)&& createDepthTexture;//清除临时RTif (!requiresDepthTexture && !createDepthTexture && !requiresDepthCopyPass && !requiresDepthPrepass){m_CameraDepthAttachment?.Release();m_ActiveCameraDepthAttachment?.Release();m_DepthTexture?.Release();}if (!createColorTexture){m_ActiveCameraColorAttachment?.Release();m_ColorBufferSystem.Dispose();}if (!copyColorPass){m_OpaqueColor?.Release();}if ((DebugHandler != null) && DebugHandler.IsActiveForCamera(ref cameraData)){DebugHandler.TryGetFullscreenDebugMode(out var fullScreenMode);if (fullScreenMode == DebugFullScreenMode.Depth){requiresDepthPrepass = true;}if (!DebugHandler.IsLightingActive){mainLightShadows = false;additionalLightShadows = false;if (!isSceneViewOrPreviewCamera){requiresDepthPrepass = false;useDepthPriming = false;generateColorGradingLUT = false;copyColorPass = false;requiresDepthCopyPass = false;}}if (useRenderPassEnabled)useRenderPassEnabled = DebugHandler.IsRenderPassSupported;}cameraData.renderer.useDepthPriming = useDepthPriming;if (this.renderingModeActual == RenderingMode.Deferred){if (m_DeferredLights.UseRenderPass && (RenderPassEvent.AfterRenderingGbuffer == renderPassInputs.requiresDepthNormalAtEvent || !useRenderPassEnabled))m_DeferredLights.DisableFramebufferFetchInput();}// Allocate m_DepthTexture if usedif ((this.renderingModeActual == RenderingMode.Deferred && !this.useRenderPassEnabled) || requiresDepthPrepass || requiresDepthCopyPass){var depthDescriptor = cameraTargetDescriptor;if ((requiresDepthPrepass && this.renderingModeActual != RenderingMode.Deferred) || !RenderingUtils.SupportsGraphicsFormat(GraphicsFormat.R32_SFloat, FormatUsage.Render)){// Debug.Log("Pass 1");depthDescriptor.graphicsFormat = GraphicsFormat.None;depthDescriptor.depthStencilFormat = k_DepthStencilFormat;depthDescriptor.depthBufferBits = k_DepthBufferBits;
#if UNITY_IOS || UNITY_ANDROID || UNITY_EDITORdepthDescriptor.width = (int)(depthDescriptor.width * 0.5);depthDescriptor.height = (int)(depthDescriptor.height * 0.5);
#endif}else{//修改CameraDepth精度和大小
#if UNITY_IOS || UNITY_ANDROID || UNITY_EDITORif (RenderingUtils.SupportsGraphicsFormat(GraphicsFormat.R16_SNorm, FormatUsage.Render))depthDescriptor.graphicsFormat = GraphicsFormat.R16_SNorm;else if (RenderingUtils.SupportsGraphicsFormat(GraphicsFormat.R16_SFloat, FormatUsage.Render))depthDescriptor.graphicsFormat = GraphicsFormat.R16_SFloat;else if (RenderingUtils.SupportsGraphicsFormat(GraphicsFormat.R16_UNorm, FormatUsage.Render))depthDescriptor.graphicsFormat = GraphicsFormat.R16_UNorm;elsedepthDescriptor.graphicsFormat = GraphicsFormat.R32_SFloat;depthDescriptor.width = (int)(depthDescriptor.width * 0.5);depthDescriptor.height = (int)(depthDescriptor.height * 0.5);
#elsedepthDescriptor.graphicsFormat = GraphicsFormat.R32_SFloat;
#endifdepthDescriptor.depthStencilFormat = GraphicsFormat.None;depthDescriptor.depthBufferBits = 0;}// Debug.Log("Pass 2" + depthDescriptor.graphicsFormat + " " + depthDescriptor.width + "  " + depthDescriptor.height);depthDescriptor.msaaSamples = 1;// Depth-Only pass don't use MSAARenderingUtils.ReAllocateIfNeeded(ref m_DepthTexture, depthDescriptor, FilterMode.Point, wrapMode: TextureWrapMode.Clamp, name: "_CameraDepthTexture");cmd.SetGlobalTexture(m_DepthTexture.name, m_DepthTexture.nameID);context.ExecuteCommandBuffer(cmd);cmd.Clear();}if (requiresRenderingLayer || (renderingModeActual == RenderingMode.Deferred && m_DeferredLights.UseRenderingLayers)){ref var renderingLayersTexture = ref m_DecalLayersTexture;string renderingLayersTextureName = "_CameraRenderingLayersTexture";if (this.renderingModeActual == RenderingMode.Deferred && m_DeferredLights.UseRenderingLayers){renderingLayersTexture = ref m_DeferredLights.GbufferAttachments[(int)m_DeferredLights.GBufferRenderingLayers];renderingLayersTextureName = renderingLayersTexture.name;}var renderingLayersDescriptor = cameraTargetDescriptor;renderingLayersDescriptor.depthBufferBits = 0;// Never have MSAA on this depth texture. When doing MSAA depth priming this is the texture that is resolved to and used for post-processing.if (!renderingLayerProvidesRenderObjectPass)renderingLayersDescriptor.msaaSamples = 1;// Depth-Only pass don't use MSAA// Find compatible render-target format for storing normals.// Shader code outputs normals in signed format to be compatible with deferred gbuffer layout.// Deferred gbuffer format is signed so that normals can be blended for terrain geometry.if (this.renderingModeActual == RenderingMode.Deferred && m_DeferredLights.UseRenderingLayers)renderingLayersDescriptor.graphicsFormat = m_DeferredLights.GetGBufferFormat(m_DeferredLights.GBufferRenderingLayers); // the one used by the gbuffer.elserenderingLayersDescriptor.graphicsFormat = RenderingLayerUtils.GetFormat(renderingLayerMaskSize);if (renderingModeActual == RenderingMode.Deferred && m_DeferredLights.UseRenderingLayers){m_DeferredLights.ReAllocateGBufferIfNeeded(renderingLayersDescriptor, (int)m_DeferredLights.GBufferRenderingLayers);}else{RenderingUtils.ReAllocateIfNeeded(ref renderingLayersTexture, renderingLayersDescriptor, FilterMode.Point, TextureWrapMode.Clamp, name: renderingLayersTextureName);}cmd.SetGlobalTexture(renderingLayersTexture.name, renderingLayersTexture.nameID);RenderingLayerUtils.SetupProperties(cmd, renderingLayerMaskSize);if (this.renderingModeActual == RenderingMode.Deferred) // As this is requested by render pass we still want to set itcmd.SetGlobalTexture("_CameraRenderingLayersTexture", renderingLayersTexture.nameID);context.ExecuteCommandBuffer(cmd);cmd.Clear();}// Allocate normal texture if usedif (requiresDepthPrepass && renderPassInputs.requiresNormalsTexture){ref var normalsTexture = ref m_NormalsTexture;string normalsTextureName = "_CameraNormalsTexture";if (this.renderingModeActual == RenderingMode.Deferred){normalsTexture = ref m_DeferredLights.GbufferAttachments[(int)m_DeferredLights.GBufferNormalSmoothnessIndex];normalsTextureName = normalsTexture.name;}var normalDescriptor = cameraTargetDescriptor;normalDescriptor.depthBufferBits = 0;// Never have MSAA on this depth texture. When doing MSAA depth priming this is the texture that is resolved to and used for post-processing.normalDescriptor.msaaSamples = useDepthPriming ? cameraTargetDescriptor.msaaSamples : 1;// Depth-Only passes don't use MSAA, unless depth priming is enabled// Find compatible render-target format for storing normals.// Shader code outputs normals in signed format to be compatible with deferred gbuffer layout.// Deferred gbuffer format is signed so that normals can be blended for terrain geometry.if (this.renderingModeActual == RenderingMode.Deferred)normalDescriptor.graphicsFormat = m_DeferredLights.GetGBufferFormat(m_DeferredLights.GBufferNormalSmoothnessIndex); // the one used by the gbuffer.elsenormalDescriptor.graphicsFormat = DepthNormalOnlyPass.GetGraphicsFormat();if (this.renderingModeActual == RenderingMode.Deferred){m_DeferredLights.ReAllocateGBufferIfNeeded(normalDescriptor, (int)m_DeferredLights.GBufferNormalSmoothnessIndex);}else{RenderingUtils.ReAllocateIfNeeded(ref normalsTexture, normalDescriptor, FilterMode.Point, TextureWrapMode.Clamp, name: normalsTextureName);}cmd.SetGlobalTexture(normalsTexture.name, normalsTexture.nameID);if (this.renderingModeActual == RenderingMode.Deferred) // As this is requested by render pass we still want to set itcmd.SetGlobalTexture("_CameraNormalsTexture", normalsTexture.nameID);context.ExecuteCommandBuffer(cmd);cmd.Clear();}if (requiresDepthPrepass){if (renderPassInputs.requiresNormalsTexture){if (this.renderingModeActual == RenderingMode.Deferred){// In deferred mode, depth-normal prepass does really primes the depth and normal buffers, instead of creating a copy.// It is necessary because we need to render depth&normal for forward-only geometry and it is the only way// to get them before the SSAO pass.int gbufferNormalIndex = m_DeferredLights.GBufferNormalSmoothnessIndex;if (m_DeferredLights.UseRenderingLayers)m_DepthNormalPrepass.Setup(m_ActiveCameraDepthAttachment, m_DeferredLights.GbufferAttachments[gbufferNormalIndex], m_DeferredLights.GbufferAttachments[m_DeferredLights.GBufferRenderingLayers]);else if (renderingLayerProvidesByDepthNormalPass)m_DepthNormalPrepass.Setup(m_ActiveCameraDepthAttachment, m_DeferredLights.GbufferAttachments[gbufferNormalIndex], m_DecalLayersTexture);elsem_DepthNormalPrepass.Setup(m_ActiveCameraDepthAttachment, m_DeferredLights.GbufferAttachments[gbufferNormalIndex]);// Only render forward-only geometry, as standard geometry will be rendered as normal into the gbuffer.if (RenderPassEvent.AfterRenderingGbuffer <= renderPassInputs.requiresDepthNormalAtEvent &&renderPassInputs.requiresDepthNormalAtEvent <= RenderPassEvent.BeforeRenderingOpaques)m_DepthNormalPrepass.shaderTagIds = k_DepthNormalsOnly;}else{if (renderingLayerProvidesByDepthNormalPass)m_DepthNormalPrepass.Setup(m_DepthTexture, m_NormalsTexture, m_DecalLayersTexture);elsem_DepthNormalPrepass.Setup(m_DepthTexture, m_NormalsTexture);}EnqueuePass(m_DepthNormalPrepass);}else{// Deferred renderer does not require a depth-prepass to generate samplable depth texture.if (this.renderingModeActual != RenderingMode.Deferred){m_DepthPrepass.Setup(cameraTargetDescriptor, m_DepthTexture);EnqueuePass(m_DepthPrepass);}}}// depth priming still needs to copy depth because the prepass doesn't target anymore CameraDepthTexture// TODO: this is unoptimal, investigate optimizationsif (useDepthPriming){m_PrimedDepthCopyPass.Setup(m_ActiveCameraDepthAttachment, m_DepthTexture);EnqueuePass(m_PrimedDepthCopyPass);}if (generateColorGradingLUT){colorGradingLutPass.ConfigureDescriptor(in renderingData.postProcessingData, out var desc, out var filterMode);RenderingUtils.ReAllocateIfNeeded(ref m_PostProcessPasses.m_ColorGradingLut, desc, filterMode, TextureWrapMode.Clamp, anisoLevel: 0, name: "_InternalGradingLut");colorGradingLutPass.Setup(colorGradingLut);EnqueuePass(colorGradingLutPass);}#if ENABLE_VR && ENABLE_XR_MODULEif (cameraData.xr.hasValidOcclusionMesh)EnqueuePass(m_XROcclusionMeshPass);
#endifbool lastCameraInTheStack = cameraData.resolveFinalTarget;if (this.renderingModeActual == RenderingMode.Deferred){if (m_DeferredLights.UseRenderPass && (RenderPassEvent.AfterRenderingGbuffer == renderPassInputs.requiresDepthNormalAtEvent || !useRenderPassEnabled))m_DeferredLights.DisableFramebufferFetchInput();EnqueueDeferred(ref renderingData, requiresDepthPrepass, renderPassInputs.requiresNormalsTexture, renderingLayerProvidesByDepthNormalPass, mainLightShadows, additionalLightShadows);}else{// Optimized store actions are very important on tile based GPUs and have a great impact on performance.// if MSAA is enabled and any of the following passes need a copy of the color or depth target, make sure the MSAA'd surface is stored// if following passes won't use it then just resolve (the Resolve action will still store the resolved surface, but discard the MSAA'd surface, which is very expensive to store).RenderBufferStoreAction opaquePassColorStoreAction = RenderBufferStoreAction.Store;if (cameraTargetDescriptor.msaaSamples > 1)opaquePassColorStoreAction = copyColorPass ? RenderBufferStoreAction.StoreAndResolve : RenderBufferStoreAction.Store;// make sure we store the depth only if following passes need it.RenderBufferStoreAction opaquePassDepthStoreAction = (copyColorPass || requiresDepthCopyPass || !lastCameraInTheStack) ? RenderBufferStoreAction.Store : RenderBufferStoreAction.DontCare;
#if ENABLE_VR && ENABLE_XR_MODULEif (cameraData.xr.enabled && cameraData.xr.copyDepth){opaquePassDepthStoreAction = RenderBufferStoreAction.Store;}
#endif// handle multisample depth resolve by setting the appropriate store actions if supportedif (requiresDepthCopyPass && cameraTargetDescriptor.msaaSamples > 1 && RenderingUtils.MultisampleDepthResolveSupported()){bool isCopyDepthAfterTransparent = m_CopyDepthPass.renderPassEvent == RenderPassEvent.AfterRenderingTransparents;// we could StoreAndResolve when the depth copy is after opaque, but performance wise doing StoreAndResolve of depth targets is more expensive than a simple Store + following depth copy pass on Apple GPUs,// because of the extra resolve step. So, unless we are copying the depth after the transparent pass, just Store the depth target.if (isCopyDepthAfterTransparent && !copyColorPass){if (opaquePassDepthStoreAction == RenderBufferStoreAction.Store)opaquePassDepthStoreAction = RenderBufferStoreAction.StoreAndResolve;else if (opaquePassDepthStoreAction == RenderBufferStoreAction.DontCare)opaquePassDepthStoreAction = RenderBufferStoreAction.Resolve;}}DrawObjectsPass renderOpaqueForwardPass = null;if (renderingLayerProvidesRenderObjectPass){renderOpaqueForwardPass = m_RenderOpaqueForwardWithRenderingLayersPass;m_RenderOpaqueForwardWithRenderingLayersPass.Setup(m_ActiveCameraColorAttachment, m_DecalLayersTexture, m_ActiveCameraDepthAttachment);}elserenderOpaqueForwardPass = m_RenderOpaqueForwardPass;renderOpaqueForwardPass.ConfigureColorStoreAction(opaquePassColorStoreAction);renderOpaqueForwardPass.ConfigureDepthStoreAction(opaquePassDepthStoreAction);// If there is any custom render pass renders to opaque pass' target before opaque pass,// we can't clear color as it contains the valid rendering output.bool hasPassesBeforeOpaque = activeRenderPassQueue.Find(x => (x.renderPassEvent <= RenderPassEvent.BeforeRenderingOpaques) && !x.overrideCameraTarget) != null;ClearFlag opaqueForwardPassClearFlag = (hasPassesBeforeOpaque || cameraData.renderType != CameraRenderType.Base)? ClearFlag.None: ClearFlag.Color;
#if ENABLE_VR && ENABLE_XR_MODULE// workaround for DX11 and DX12 XR test failures.// XRTODO: investigate DX XR clear issues.if (SystemInfo.usesLoadStoreActions)
#endifrenderOpaqueForwardPass.ConfigureClear(opaqueForwardPassClearFlag, Color.black);EnqueuePass(renderOpaqueForwardPass);}if (camera.clearFlags == CameraClearFlags.Skybox && cameraData.renderType != CameraRenderType.Overlay){if (RenderSettings.skybox != null || (camera.TryGetComponent(out Skybox cameraSkybox) && cameraSkybox.material != null))EnqueuePass(m_DrawSkyboxPass);}// If a depth texture was created we necessarily need to copy it, otherwise we could have render it to a renderbuffer.// Also skip if Deferred+RenderPass as CameraDepthTexture is used and filled by the GBufferPass// however we might need the depth texture with Forward-only pass rendered to it, so enable the copy depth in that caseif (requiresDepthCopyPass && !(this.renderingModeActual == RenderingMode.Deferred && useRenderPassEnabled && !renderPassInputs.requiresDepthTexture)){m_CopyDepthPass.Setup(m_ActiveCameraDepthAttachment, m_DepthTexture);EnqueuePass(m_CopyDepthPass);}// Set the depth texture to the far Z if we do not have a depth prepass or copy depth// Don't do this for Overlay cameras to not lose depth data in between cameras (as Base is guaranteed to be first)if (cameraData.renderType == CameraRenderType.Base && !requiresDepthPrepass && !requiresDepthCopyPass)Shader.SetGlobalTexture("_CameraDepthTexture", SystemInfo.usesReversedZBuffer ? Texture2D.blackTexture : Texture2D.whiteTexture);if (copyColorPass){// TODO: Downsampling method should be stored in the renderer instead of in the asset.// We need to migrate this data to renderer. For now, we query the method in the active asset.Downsampling downsamplingMethod = UniversalRenderPipeline.asset.opaqueDownsampling;var descriptor = cameraTargetDescriptor;CopyColorPass.ConfigureDescriptor(downsamplingMethod, ref descriptor, out var filterMode);RenderingUtils.ReAllocateIfNeeded(ref m_OpaqueColor, descriptor, filterMode, TextureWrapMode.Clamp, name: "_CameraOpaqueTexture");m_CopyColorPass.Setup(m_ActiveCameraColorAttachment, m_OpaqueColor, downsamplingMethod);EnqueuePass(m_CopyColorPass);}// Motion vectorsif (renderPassInputs.requiresMotionVectors){var colorDesc = cameraTargetDescriptor;colorDesc.graphicsFormat = MotionVectorRenderPass.k_TargetFormat;colorDesc.depthBufferBits = (int)DepthBits.None;colorDesc.msaaSamples = 1;  // Disable MSAA, consider a pixel resolve for half left velocity and half right velocity --> no velocity, which is untrue.RenderingUtils.ReAllocateIfNeeded(ref m_MotionVectorColor, colorDesc, FilterMode.Point, TextureWrapMode.Clamp, name: "_MotionVectorTexture");var depthDescriptor = cameraTargetDescriptor;depthDescriptor.graphicsFormat = GraphicsFormat.None;depthDescriptor.msaaSamples = 1;RenderingUtils.ReAllocateIfNeeded(ref m_MotionVectorDepth, depthDescriptor, FilterMode.Point, TextureWrapMode.Clamp, name: "_MotionVectorDepthTexture");m_MotionVectorPass.Setup(m_MotionVectorColor, m_MotionVectorDepth);EnqueuePass(m_MotionVectorPass);}#if ADAPTIVE_PERFORMANCE_2_1_0_OR_NEWERif (needTransparencyPass)
#endif{if (transparentsNeedSettingsPass){EnqueuePass(m_TransparentSettingsPass);}// if this is not lastCameraInTheStack we still need to Store, since the MSAA buffer might be needed by the Overlay camerasRenderBufferStoreAction transparentPassColorStoreAction = cameraTargetDescriptor.msaaSamples > 1 && lastCameraInTheStack ? RenderBufferStoreAction.Resolve : RenderBufferStoreAction.Store;RenderBufferStoreAction transparentPassDepthStoreAction = lastCameraInTheStack ? RenderBufferStoreAction.DontCare : RenderBufferStoreAction.Store;// If CopyDepthPass pass event is scheduled on or after AfterRenderingTransparent, we will need to store the depth buffer or resolve (store for now until latest trunk has depth resolve support) it for MSAA caseif (requiresDepthCopyPass && m_CopyDepthPass.renderPassEvent >= RenderPassEvent.AfterRenderingTransparents){transparentPassDepthStoreAction = RenderBufferStoreAction.Store;// handle depth resolve on platforms supporting itif (cameraTargetDescriptor.msaaSamples > 1 && RenderingUtils.MultisampleDepthResolveSupported())transparentPassDepthStoreAction = RenderBufferStoreAction.Resolve;}m_RenderTransparentForwardPass.ConfigureColorStoreAction(transparentPassColorStoreAction);m_RenderTransparentForwardPass.ConfigureDepthStoreAction(transparentPassDepthStoreAction);EnqueuePass(m_RenderTransparentForwardPass);}EnqueuePass(m_OnRenderObjectCallbackPass);bool shouldRenderUI = cameraData.rendersOverlayUI;bool outputToHDR = cameraData.isHDROutputActive;if (shouldRenderUI && outputToHDR){m_DrawOffscreenUIPass.Setup(ref cameraData, k_DepthBufferBits);EnqueuePass(m_DrawOffscreenUIPass);}bool hasCaptureActions = renderingData.cameraData.captureActions != null && lastCameraInTheStack;// When FXAA or scaling is active, we must perform an additional pass at the end of the frame for the following reasons:// 1. FXAA expects to be the last shader running on the image before it's presented to the screen. Since users are allowed//    to add additional render passes after post processing occurs, we can't run FXAA until all of those passes complete as well.//    The FinalPost pass is guaranteed to execute after user authored passes so FXAA is always run inside of it.// 2. UberPost can only handle upscaling with linear filtering. All other filtering methods require the FinalPost pass.// 3. TAA sharpening using standalone RCAS pass is required. (When upscaling is not enabled).bool applyFinalPostProcessing = anyPostProcessing && lastCameraInTheStack &&((renderingData.cameraData.antialiasing == AntialiasingMode.FastApproximateAntialiasing) ||((renderingData.cameraData.imageScalingMode == ImageScalingMode.Upscaling) && (renderingData.cameraData.upscalingFilter != ImageUpscalingFilter.Linear)) ||(renderingData.cameraData.IsTemporalAAEnabled() && renderingData.cameraData.taaSettings.contrastAdaptiveSharpening > 0.0f));// When post-processing is enabled we can use the stack to resolve rendering to camera target (screen or RT).// However when there are render passes executing after post we avoid resolving to screen so rendering continues (before sRGBConversion etc)bool resolvePostProcessingToCameraTarget = !hasCaptureActions && !hasPassesAfterPostProcessing && !applyFinalPostProcessing;bool needsColorEncoding = DebugHandler == null || !DebugHandler.HDRDebugViewIsActive(ref cameraData);//处理后处理渲染没有组件临时RT释放if (applyPostProcessing){if (clearInPostProcess){m_PostProcessPasses.m_AfterPostProcessColor?.Release();postProcessPass?.Dispose();}else{var desc = PostProcessPass.GetCompatibleDescriptor(cameraTargetDescriptor, cameraTargetDescriptor.width, cameraTargetDescriptor.height, cameraTargetDescriptor.graphicsFormat, DepthBits.None);RenderingUtils.ReAllocateIfNeeded(ref m_PostProcessPasses.m_AfterPostProcessColor, desc, FilterMode.Point, TextureWrapMode.Clamp, name: "_AfterPostProcessTexture");}}else{m_PostProcessPasses.m_AfterPostProcessColor?.Release();postProcessPass?.Dispose();}if (lastCameraInTheStack){SetupFinalPassDebug(ref cameraData);// Post-processing will resolve to final target. No need for final blit pass.if (applyPostProcessing){// if resolving to screen we need to be able to perform sRGBConversion in post-processing if necessarybool doSRGBEncoding = resolvePostProcessingToCameraTarget && needsColorEncoding;postProcessPass.Setup(cameraTargetDescriptor, m_ActiveCameraColorAttachment, resolvePostProcessingToCameraTarget, m_ActiveCameraDepthAttachment, colorGradingLut, m_MotionVectorColor, applyFinalPostProcessing, doSRGBEncoding);EnqueuePass(postProcessPass);}var sourceForFinalPass = m_ActiveCameraColorAttachment;// Do FXAA or any other final post-processing effect that might need to run after AA.if (applyFinalPostProcessing){finalPostProcessPass.SetupFinalPass(sourceForFinalPass, true, needsColorEncoding);EnqueuePass(finalPostProcessPass);}if (renderingData.cameraData.captureActions != null){EnqueuePass(m_CapturePass);}// if post-processing then we already resolved to camera target while doing post.// Also only do final blit if camera is not rendering to RT.bool cameraTargetResolved =// final PP always blit to camera targetapplyFinalPostProcessing ||// no final PP but we have PP stack. In that case it blit unless there are render pass after PP(applyPostProcessing && !hasPassesAfterPostProcessing && !hasCaptureActions) ||// offscreen camera rendering to a texture, we don't need a blit pass to resolve to screenm_ActiveCameraColorAttachment.nameID == m_XRTargetHandleAlias.nameID;// We need final blit to resolve to screenif (!cameraTargetResolved){m_FinalBlitPass.Setup(cameraTargetDescriptor, sourceForFinalPass);EnqueuePass(m_FinalBlitPass);}if (shouldRenderUI && !outputToHDR){EnqueuePass(m_DrawOverlayUIPass);}#if ENABLE_VR && ENABLE_XR_MODULEif (cameraData.xr.enabled){// active depth is depth target, we don't need a blit pass to resolvebool depthTargetResolved = m_ActiveCameraDepthAttachment.nameID == cameraData.xr.renderTarget;if (!depthTargetResolved && cameraData.xr.copyDepth){m_XRCopyDepthPass.Setup(m_ActiveCameraDepthAttachment, m_XRTargetHandleAlias);m_XRCopyDepthPass.CopyToDepth = true;EnqueuePass(m_XRCopyDepthPass);}}
#endif}// stay in RT so we resume rendering on stack after post-processingelse if (applyPostProcessing){postProcessPass.Setup(cameraTargetDescriptor, m_ActiveCameraColorAttachment, false, m_ActiveCameraDepthAttachment, colorGradingLut, m_MotionVectorColor, false, false);EnqueuePass(postProcessPass);}#if UNITY_EDITORif (isSceneViewOrPreviewCamera || (isGizmosEnabled && lastCameraInTheStack)){// Scene view camera should always resolve target (not stacked)m_FinalDepthCopyPass.Setup(m_DepthTexture, k_CameraTarget);m_FinalDepthCopyPass.CopyToDepth = true;m_FinalDepthCopyPass.MssaSamples = 0;// Turning off unnecessary NRP in Editor because of MSAA mistmatch between CameraTargetDescriptor vs camera backbuffer// NRP layer considers this being a pass with MSAA samples by checking CameraTargetDescriptor taken from RP asset// while the camera backbuffer has a single samplem_FinalDepthCopyPass.useNativeRenderPass = false;EnqueuePass(m_FinalDepthCopyPass);}
#endif}/// <inheritdoc />public override void SetupLights(ScriptableRenderContext context, ref RenderingData renderingData){m_ForwardLights.Setup(context, ref renderingData);if (this.renderingModeActual == RenderingMode.Deferred)m_DeferredLights.SetupLights(context, ref renderingData);}/// <inheritdoc />public override void SetupCullingParameters(ref ScriptableCullingParameters cullingParameters,ref CameraData cameraData){// TODO: PerObjectCulling also affect reflection probes. Enabling it for now.// if (asset.additionalLightsRenderingMode == LightRenderingMode.Disabled ||//     asset.maxAdditionalLightsCount == 0)if (renderingModeActual == RenderingMode.ForwardPlus){cullingParameters.cullingOptions |= CullingOptions.DisablePerObjectCulling;}// We disable shadow casters if both shadow casting modes are turned off// or the shadow distance has been turned down to zerobool isShadowCastingDisabled = !UniversalRenderPipeline.asset.supportsMainLightShadows && !UniversalRenderPipeline.asset.supportsAdditionalLightShadows;bool isShadowDistanceZero = Mathf.Approximately(cameraData.maxShadowDistance, 0.0f);if (isShadowCastingDisabled || isShadowDistanceZero){cullingParameters.cullingOptions &= ~CullingOptions.ShadowCasters;}if (this.renderingModeActual == RenderingMode.Deferred)cullingParameters.maximumVisibleLights = 0xFFFF;else if (this.renderingModeActual == RenderingMode.ForwardPlus){// We don't add one to the maximum light because mainlight is treated as any other light.cullingParameters.maximumVisibleLights = UniversalRenderPipeline.maxVisibleAdditionalLights;// Do not sort reflection probe from engine it will come in reverse order from what we need.cullingParameters.reflectionProbeSortingCriteria = ReflectionProbeSortingCriteria.None;}else{// We set the number of maximum visible lights allowed and we add one for the mainlight...//// Note: However ScriptableRenderContext.Cull() does not differentiate between light types.//       If there is no active main light in the scene, ScriptableRenderContext.Cull() might return  ( cullingParameters.maximumVisibleLights )  visible additional lights.//       i.e ScriptableRenderContext.Cull() might return  ( UniversalRenderPipeline.maxVisibleAdditionalLights + 1 )  visible additional lights !cullingParameters.maximumVisibleLights = UniversalRenderPipeline.maxVisibleAdditionalLights + 1;}cullingParameters.shadowDistance = cameraData.maxShadowDistance;cullingParameters.conservativeEnclosingSphere = UniversalRenderPipeline.asset.conservativeEnclosingSphere;cullingParameters.numIterationsEnclosingSphere = UniversalRenderPipeline.asset.numIterationsEnclosingSphere;}/// <inheritdoc />public override void FinishRendering(CommandBuffer cmd){m_ColorBufferSystem.Clear();m_ActiveCameraColorAttachment = null;m_ActiveCameraDepthAttachment = null;}void EnqueueDeferred(ref RenderingData renderingData, bool hasDepthPrepass, bool hasNormalPrepass, bool hasRenderingLayerPrepass, bool applyMainShadow, bool applyAdditionalShadow){m_DeferredLights.Setup(ref renderingData,applyAdditionalShadow ? m_AdditionalLightsShadowCasterPass : null,hasDepthPrepass,hasNormalPrepass,hasRenderingLayerPrepass,m_DepthTexture,m_ActiveCameraDepthAttachment,m_ActiveCameraColorAttachment);// Need to call Configure for both of these passes to setup input attachments as first frame otherwise will raise errorsif (useRenderPassEnabled && m_DeferredLights.UseRenderPass){m_GBufferPass.Configure(null, renderingData.cameraData.cameraTargetDescriptor);m_DeferredPass.Configure(null, renderingData.cameraData.cameraTargetDescriptor);}EnqueuePass(m_GBufferPass);//Must copy depth for deferred shading: TODO wait for API fix to bind depth texture as read-only resource.if (!useRenderPassEnabled || !m_DeferredLights.UseRenderPass){m_GBufferCopyDepthPass.Setup(m_CameraDepthAttachment, m_DepthTexture);EnqueuePass(m_GBufferCopyDepthPass);}EnqueuePass(m_DeferredPass);EnqueuePass(m_RenderOpaqueForwardOnlyPass);}private struct RenderPassInputSummary{internal bool requiresDepthTexture;internal bool requiresDepthPrepass;internal bool requiresNormalsTexture;internal bool requiresColorTexture;internal bool requiresColorTextureCreated;internal bool requiresMotionVectors;internal RenderPassEvent requiresDepthNormalAtEvent;internal RenderPassEvent requiresDepthTextureEarliestEvent;}private RenderPassInputSummary GetRenderPassInputs(ref RenderingData renderingData){RenderPassEvent beforeMainRenderingEvent = m_RenderingMode == RenderingMode.Deferred ? RenderPassEvent.BeforeRenderingGbuffer : RenderPassEvent.BeforeRenderingOpaques;RenderPassInputSummary inputSummary = new RenderPassInputSummary();inputSummary.requiresDepthNormalAtEvent = RenderPassEvent.BeforeRenderingOpaques;inputSummary.requiresDepthTextureEarliestEvent = RenderPassEvent.BeforeRenderingPostProcessing;for (int i = 0; i < activeRenderPassQueue.Count; ++i){ScriptableRenderPass pass = activeRenderPassQueue[i];bool needsDepth = (pass.input & ScriptableRenderPassInput.Depth) != ScriptableRenderPassInput.None;bool needsNormals = (pass.input & ScriptableRenderPassInput.Normal) != ScriptableRenderPassInput.None;bool needsColor = (pass.input & ScriptableRenderPassInput.Color) != ScriptableRenderPassInput.None;bool needsMotion = (pass.input & ScriptableRenderPassInput.Motion) != ScriptableRenderPassInput.None;bool eventBeforeMainRendering = pass.renderPassEvent <= beforeMainRenderingEvent;// TODO: Need a better way to handle this, probably worth to recheck after render graph// DBuffer requires color texture created as it does not handle y flip correctlyif (pass is DBufferRenderPass dBufferRenderPass){inputSummary.requiresColorTextureCreated = true;}inputSummary.requiresDepthTexture |= needsDepth;inputSummary.requiresDepthPrepass |= needsNormals || needsDepth && eventBeforeMainRendering;inputSummary.requiresNormalsTexture |= needsNormals;inputSummary.requiresColorTexture |= needsColor;inputSummary.requiresMotionVectors |= needsMotion;if (needsDepth)inputSummary.requiresDepthTextureEarliestEvent = (RenderPassEvent)Mathf.Min((int)pass.renderPassEvent, (int)inputSummary.requiresDepthTextureEarliestEvent);if (needsNormals || needsDepth)inputSummary.requiresDepthNormalAtEvent = (RenderPassEvent)Mathf.Min((int)pass.renderPassEvent, (int)inputSummary.requiresDepthNormalAtEvent);}// NOTE: TAA and motion vector dependencies added here to share between Execute and Render (Graph) paths.// TAA in postprocess requires motion to function.if (renderingData.cameraData.IsTemporalAAEnabled())inputSummary.requiresMotionVectors = true;// Motion vectors imply depthif (inputSummary.requiresMotionVectors){inputSummary.requiresDepthTexture = true;inputSummary.requiresDepthTextureEarliestEvent = (RenderPassEvent)Mathf.Min((int)m_MotionVectorPass.renderPassEvent, (int)inputSummary.requiresDepthTextureEarliestEvent);}return inputSummary;}void CreateCameraRenderTarget(ScriptableRenderContext context, ref RenderTextureDescriptor descriptor, bool primedDepth, CommandBuffer cmd, ref CameraData cameraData){using (new ProfilingScope(null, Profiling.createCameraRenderTarget)){if (m_ColorBufferSystem.PeekBackBuffer() == null || m_ColorBufferSystem.PeekBackBuffer().nameID != BuiltinRenderTextureType.CameraTarget){m_ActiveCameraColorAttachment = m_ColorBufferSystem.GetBackBuffer(cmd);ConfigureCameraColorTarget(m_ActiveCameraColorAttachment);cmd.SetGlobalTexture("_CameraColorTexture", m_ActiveCameraColorAttachment.nameID);//Set _AfterPostProcessTexture, users might still rely on this although it is now always the cameratarget due to swapbuffercmd.SetGlobalTexture("_AfterPostProcessTexture", m_ActiveCameraColorAttachment.nameID);}if (m_CameraDepthAttachment == null || m_CameraDepthAttachment.nameID != BuiltinRenderTextureType.CameraTarget){//清除未旧附件m_CameraDepthAttachment?.Release();var depthDescriptor = descriptor;depthDescriptor.useMipMap = false;depthDescriptor.autoGenerateMips = false;depthDescriptor.bindMS = false;bool hasMSAA = depthDescriptor.msaaSamples > 1 && (SystemInfo.supportsMultisampledTextures != 0);// if MSAA is enabled and we are not resolving depth, which we only do if the CopyDepthPass is AfterTransparents,// then we want to bind the multisampled surface.if (hasMSAA){// if depth priming is enabled the copy depth primed pass is meant to do the MSAA resolve, so we want to bind the MS surfaceif (IsDepthPrimingEnabled(ref cameraData))depthDescriptor.bindMS = true;elsedepthDescriptor.bindMS = !(RenderingUtils.MultisampleDepthResolveSupported() &&SystemInfo.supportsMultisampleAutoResolve &&m_CopyDepthMode == CopyDepthMode.AfterTransparents);}// binding MS surfaces is not supported by the GLES backend, and it won't be fixed after investigating// the high performance impact of potential fixes, which would make it more expensive than depth prepass (fogbugz 1339401 for more info)if (IsGLESDevice())depthDescriptor.bindMS = false;depthDescriptor.graphicsFormat = GraphicsFormat.None;depthDescriptor.depthStencilFormat = k_DepthStencilFormat;//设置CameraDepthAttachment精度if (!cameraData.camera.orthographic)depthDescriptor.depthBufferBits = k_DepthBufferBits;elsedepthDescriptor.depthBufferBits = 24;RenderingUtils.ReAllocateIfNeeded(ref m_CameraDepthAttachment, depthDescriptor, FilterMode.Point, TextureWrapMode.Clamp, name: "_CameraDepthAttachment");cmd.SetGlobalTexture(m_CameraDepthAttachment.name, m_CameraDepthAttachment.nameID);// update the descriptor to match the depth attachmentdescriptor.depthStencilFormat = depthDescriptor.depthStencilFormat;descriptor.depthBufferBits = depthDescriptor.depthBufferBits;}}context.ExecuteCommandBuffer(cmd);cmd.Clear();}bool PlatformRequiresExplicitMsaaResolve(){
#if UNITY_EDITOR// In the editor play-mode we use a Game View Render Texture, with// samples count forced to 1 so we always need to do an explicit MSAA resolve.return true;
#else// On Metal/iOS the MSAA resolve is done implicitly as part of the renderpass, so we do not need an extra intermediate pass for the explicit autoresolve.// Note: On Vulkan Standalone, despite SystemInfo.supportsMultisampleAutoResolve being true, the backbuffer has only 1 sample, so we still require// the explicit resolve on non-mobile platforms with supportsMultisampleAutoResolve.return !(SystemInfo.supportsMultisampleAutoResolve && Application.isMobilePlatform)&& SystemInfo.graphicsDeviceType != GraphicsDeviceType.Metal;
#endif}/// <summary>/// Checks if the pipeline needs to create a intermediate render texture./// </summary>/// <param name="cameraData">CameraData contains all relevant render target information for the camera.</param>/// <seealso cref="CameraData"/>/// <returns>Return true if pipeline needs to render to a intermediate render texture.</returns>bool RequiresIntermediateColorTexture(ref CameraData cameraData){// When rendering a camera stack we always create an intermediate render texture to composite camera results.// We create it upon rendering the Base camera.if (cameraData.renderType == CameraRenderType.Base && !cameraData.resolveFinalTarget)return true;// Always force rendering into intermediate color texture if deferred rendering mode is selected.// Reason: without intermediate color texture, the target camera texture is y-flipped.// However, the target camera texture is bound during gbuffer pass and deferred pass.// Gbuffer pass will not be y-flipped because it is MRT (see ScriptableRenderContext implementation),// while deferred pass will be y-flipped, which breaks rendering.// This incurs an extra blit into at the end of rendering.if (this.renderingModeActual == RenderingMode.Deferred)return true;bool isSceneViewCamera = cameraData.isSceneViewCamera;var cameraTargetDescriptor = cameraData.cameraTargetDescriptor;int msaaSamples = cameraTargetDescriptor.msaaSamples;bool isScaledRender = cameraData.imageScalingMode != ImageScalingMode.None;bool isCompatibleBackbufferTextureDimension = cameraTargetDescriptor.dimension == TextureDimension.Tex2D;bool requiresExplicitMsaaResolve = msaaSamples > 1 && PlatformRequiresExplicitMsaaResolve();bool isOffscreenRender = cameraData.targetTexture != null && !isSceneViewCamera;bool isCapturing = cameraData.captureActions != null;#if ENABLE_VR && ENABLE_XR_MODULEif (cameraData.xr.enabled){isScaledRender = false;isCompatibleBackbufferTextureDimension = cameraData.xr.renderTargetDesc.dimension == cameraTargetDescriptor.dimension;}
#endifbool postProcessEnabled = cameraData.postProcessEnabled && m_PostProcessPasses.isCreated;bool requiresBlitForOffscreenCamera = postProcessEnabled || cameraData.requiresOpaqueTexture || requiresExplicitMsaaResolve || !cameraData.isDefaultViewport;if (isOffscreenRender)return requiresBlitForOffscreenCamera;return requiresBlitForOffscreenCamera || isSceneViewCamera || isScaledRender || cameraData.isHdrEnabled ||!isCompatibleBackbufferTextureDimension || isCapturing || cameraData.requireSrgbConversion;}bool CanCopyDepth(ref CameraData cameraData){bool msaaEnabledForCamera = cameraData.cameraTargetDescriptor.msaaSamples > 1;bool supportsTextureCopy = SystemInfo.copyTextureSupport != CopyTextureSupport.None;bool supportsDepthTarget = RenderingUtils.SupportsRenderTextureFormat(RenderTextureFormat.Depth);bool supportsDepthCopy = !msaaEnabledForCamera && (supportsDepthTarget || supportsTextureCopy);bool msaaDepthResolve = msaaEnabledForCamera && SystemInfo.supportsMultisampledTextures != 0;// copying MSAA depth on GLES3 is giving invalid results. This won't be fixed by the driver team because it would introduce performance issues (more info in the Fogbugz issue 1339401 comments)if (IsGLESDevice() && msaaDepthResolve)return false;return supportsDepthCopy || msaaDepthResolve;}internal override void SwapColorBuffer(CommandBuffer cmd){m_ColorBufferSystem.Swap();//Check if we are using the depth that is attached to color bufferif (m_ActiveCameraDepthAttachment.nameID != BuiltinRenderTextureType.CameraTarget)ConfigureCameraTarget(m_ColorBufferSystem.GetBackBuffer(cmd), m_ActiveCameraDepthAttachment);elseConfigureCameraColorTarget(m_ColorBufferSystem.GetBackBuffer(cmd));m_ActiveCameraColorAttachment = m_ColorBufferSystem.GetBackBuffer(cmd);cmd.SetGlobalTexture("_CameraColorTexture", m_ActiveCameraColorAttachment.nameID);//Set _AfterPostProcessTexture, users might still rely on this although it is now always the cameratarget due to swapbuffercmd.SetGlobalTexture("_AfterPostProcessTexture", m_ActiveCameraColorAttachment.nameID);}internal override RTHandle GetCameraColorFrontBuffer(CommandBuffer cmd){return m_ColorBufferSystem.GetFrontBuffer(cmd);}internal override RTHandle GetCameraColorBackBuffer(CommandBuffer cmd){return m_ColorBufferSystem.GetBackBuffer(cmd);}internal override void EnableSwapBufferMSAA(bool enable){m_ColorBufferSystem.EnableMSAA(enable);}}
}