Unity RenderFeature Configure和OnCameraSetup之区别

调用顺序

首先，源码描述为：

/// <summary>
/// This method is called by the renderer before rendering a camera
/// Override this method if you need to to configure render targets and their clear state, and to create temporary render target textures.
/// If a render pass doesn't override this method, this render pass renders to the active Camera's render target.
/// You should never call CommandBuffer.SetRenderTarget. Instead call <c>ConfigureTarget</c> and <c>ConfigureClear</c>.
/// </summary>
/// <param name="cmd">CommandBuffer to enqueue rendering commands. This will be executed by the pipeline.</param>
/// <param name="renderingData">Current rendering state information</param>
/// <seealso cref="ConfigureTarget"/>
/// <seealso cref="ConfigureClear"/>
public virtual void OnCameraSetup(CommandBuffer cmd, ref RenderingData renderingData)
{ }

/// <summary>
/// This method is called by the renderer before executing the render pass.
/// Override this method if you need to to configure render targets and their clear state, and to create temporary render target textures.
/// If a render pass doesn't override this method, this render pass renders to the active Camera's render target.
/// You should never call CommandBuffer.SetRenderTarget. Instead call <c>ConfigureTarget</c> and <c>ConfigureClear</c>.
/// </summary>
/// <param name="cmd">CommandBuffer to enqueue rendering commands. This will be executed by the pipeline.</param>
/// <param name="cameraTextureDescriptor">Render texture descriptor of the camera render target.</param>
/// <seealso cref="ConfigureTarget"/>
/// <seealso cref="ConfigureClear"/>
public virtual void Configure(CommandBuffer cmd, RenderTextureDescriptor cameraTextureDescriptor)
{ }

实际上已经表明了其执行顺序：

OnCameraSetup(): before rendering a camera

Configure(): before executing the render pass

即Configure发生在OnCameraSetup之后。

其余注释完全相同，都建议了在其内部使用ConfigureTarget、ConfigureClear。

参数区别

 进一步的，进入UnityEngine.Rendering.Universal.ScriptableRenderer可以看到执行逻辑：

public void Execute(ScriptableRenderContext context, ref RenderingData renderingData)
{
    ...
    ref CameraData cameraData = ref renderingData.cameraData;
    ...
    if (rendererFeatures.Count != 0 && !renderingData.cameraData.isPreviewCamera)
        SetupRenderPasses(in renderingData); ⚠️调用了SetupRenderPasses
    ...

    InternalStartRendering(context, ref renderingData);ℹ️ℹ️ℹ️在此处调用了OnCameraSetup
    ...
    {
        ...ClearRenderingState(cmd);清理相机环境
        ...更新Shader Time全局变量
    }
    ...
    对RenderPass进行排序
    SortStable(m_ActiveRenderPassQueue);
    ...
    
    foreach (var pass in activeRenderPassQueue)
        pass.Configure(cmd, cameraData.cameraTargetDescriptor);ℹ️ℹ️ℹ️此处为Configure

    ...
    SetupLights(context, ref renderingData);

    ...VFX.VFXManager...
    
    ...根据Pass的Inject时机，分阶段执行相应Pass的Execute
（执行相应Pass的Execute中安插了一次对SetPerCameraProperties的调用，用于设置相机属性，具体来说是billboard、unity_CameraWorldClipPlanes、_ProjectionParams。
之所以安插在此处执行而不是在初始化阶段执行，解释是用于VR多目适配：每个眼睛都要执行一次此设置）
}

其中InternalStartRendering的核心逻辑为：

for (int i = 0; i < m_ActiveRenderPassQueue.Count; ++i)
    m_ActiveRenderPassQueue[i].OnCameraSetup(renderingData.commandBuffer, ref renderingData);

可以得出结论：

两者参数没有本质区别，Configure的参数是OnCameraSetup的子集。

但可以看出一点的是，不同Pass间Configure的调用时刻是有序的，SortStable将Pass按照renderPassEvent的注入顺序排序。

除此之外，两者参数中的cmd也相同，都已经过Clear，同时当退出函数后，也会对其进行再一次的Clear。

如果非要说出不同的话，那就是Configure之前OnCameraSetup之后确实发生了CameraSetup：

ClearRenderingState(cmd);

在其中使用了cmd.DisableShaderKeyword设置Keyword，但这通常不会对函数体内产生什么影响：没人会在Configure、OnCameraSetup中处理这些数据。

总结

如果不关心不同Pass间的执行顺序，则Configure中的操作，完全可以在OnCameraSetup中完成。