Deferred shading. Position Conversion

[AndreyPrakhov]

In the document "Redesign of the rendering pipeline", GBuffer assumed the output of the "position". He's not here right now. I guess this was abandoned in favor of converting the pixel position from the depth buffer. I am currently working on a lighting implementation from PointLight and I am trying to use this feature in a shader.

float viewZ = getViewZ( gData.depth, cameraParams.x, cameraParams.y);
vec3 vViewPosition = getViewPosition(uv, gData.depth, viewZ,projectionMatrix,projectionMatrixInverse);

Of course, nothing happens so easily and the light source begins to wander around the stage )

I have two questions:

• What's wrong with using the getViewPosition function
• Is there a way to add a position calculation to GBuffer using normal means, rather than rewriting part of the code

[AndreyPrakhov]

Hmm, the first question is removed, I calculated vViewPosition incorrectly. If anyone needs it, it looks like this:

vec3 position = getViewPosition(uv, gData.depth, viewZ,projectionMatrix,projectionMatrixInverse);
vec4 mvPosition = viewMatrix * vec4( position, 1.0 );
vec3 vViewPosition = - mvPosition.xyz;

The question of how to add your own buffers to GBuffer remains relevant

[vanruesc]

how to add your own buffers to GBuffer

That should technically be possible by passing a GBufferConfig with additional data into GeometryPass. Rendering to the custom texture attachment would have to be implemented by modifying one of three's ShaderChunk entries. This is currently untested though.

I calculated vViewPosition incorrectly. If anyone needs it, it looks like this

Thanks for that snippet. The position data is currently missing from GData, but I'm planning on adding it as an auto-computed field similar to how GData.luminance is implemented.

[vanruesc]

I've added support for GData.POSITION in [email protected]. If an Effect specifies GBuffer.DEPTH as input and accesses gData.position, the field will be populated with the view position.

From what I've read, lighting calculations should either be done in view space or world space. If you need the world position, you can #include <pp_world_utils_pars_fragment> and then use getWorldPosition(gData.position, uv).

[vanruesc]

Also, the parameters of getViewZ and getViewPosition have changed:

• near and far have been removed from getViewZ()
• projectionMatrix and projectionMatrixInverse have been removed from getViewPosition()
  • viewZ is now optional via an overload.