Actually I might just be able to boil the number of render targets to 3, but that's the minimum, the thing is that I'm using the approach of deferred rendering, which consumes some render targets.
Btw. How on earth do you combine 6 3d cube map faces into one whole 3d cube map, never heard of this being done with 3d textures, maybe it's not possible yet...
Well, this is what I do:
...
Do the Voxelisation (done, kinda)
...
Render the scene with Vox Data, output to different RTS (Deferred Rendering).
So I'm not actually writing to the UAV's anymore, that has already been done, can I send the UAVs without being in this writing stage?
I guess I could have a separate pass, where only one color is being set in the pixel shader, the GI only color, then sample it on top to the scene.
You can make a shader resource view for your UAV buffer and use that in the shader stage as a texture input (Texture3D, etc). There's no need to use the UAVs at this point anymore. :)
