Using GrabPass to grab depth and GBuffer contents

I'd like to know how to use GrabPass to grab the contents of the depth buffer and GBuffers. The documentation specifies that it "grabs the current screen contents into a texture", however this doesn't clarify what it does when there are multiple buffers worth of screen contents, as it would require multiple textures to write into.

A simpler problem possibly to solve would be using a grab pass to copy the depth buffer so that opaque shaders can achieve depth-based effects.

Otherwise, if GrabPass cannot be used this way, is there a way to invoke a script (perhaps a command buffer) at a particular point in the Render Queue (such as "AlphaTest+48") that copies the various g/depth buffers into global shader textures? This would achieve a similar effect.

 

In deferred shading, Unity writes the g-buffer to the bound "_CameraGBufferTexture" textures, according to the documentation:

https://docs.unity3d.com/Manual/RenderTech-DeferredShading.html

Code (CSharp):

1.  
2. half4 gbuffer0 = tex2D (_CameraGBufferTexture0, uv); // Diffuse RGB, Occlusion A
3. half4 gbuffer1 = tex2D (_CameraGBufferTexture1, uv); // Specular RGB, Smoothness A
4. half4 gbuffer2 = tex2D (_CameraGBufferTexture2, uv); // Normal RGB

https://forum.unity.com/threads/accessing-deferred-renderer-g-buffers.315223/#post-2052380


Depth is stored in "_CameraDepthTexture" or "_CameraDepthNormalTexture", depending on the depthTextureMode.

 

This is correct. However, if you try accessing the depth texture while within an Opaque shader, you'll find that its contents are invalid. They may be the previous frame's depth buffer, or if you have multiple cameras, it may contain the depth contents of the last camera that rendered and thus produces and unpredictable result.

Unfortunately you cannot both read from and write to the same depth texture at the same time, even if it's just the single depth value that you're currently overwriting. It is a similar way with the color buffer; you need a grab pass to copy the contents so that you may read from the copy while writing to the original.

A grab pass for depth and G Buffers would allow one to more efficiently render deferred decals for example. Otherwise, you are required to use inefficient script callbacks and continuously write command buffers which is an inferior approach compared to just letting Unity's Renderer component render the things for you.

 

Hello zergling. Is there a chance you have figured this out?