What happens to the precomputation if e.g. we change the camera angle? If we need to redo the precomputation, is there some kind of delta mechanism to help with the process?
Could we compute blue noise in 3d space first, then use the 2d version of the distribution from the POV of our camera for our sampling?
So if it's not that expensive these days, is it much more common?
In cases of providing additional lighting to a scene with environment maps/lighting, how would a pre-computed sampling / map be interfaced with traditional path tracing methods?