All the currently implemented terrain renderers in this project are approximate. Each having its own set of compromises between the speed and quality. It would be useful to have a method as a quality baseline: something that would render ideally well, even if it's very slow.
Here is such method:
find out the XY area of the terrain that may end up affecting the screen
find out the mapping from instance ID to each and every point to this area
launch an instanced draw that would draw lines: one for the lower chunk, one for the higher chunk
With depth write enabled, and some trickery to make the front texels drawn first, this can end up in an acceptable performance.
Technically, that would be "true" Painter's algorithm, unlike the current compute scattering method.
All the currently implemented terrain renderers in this project are approximate. Each having its own set of compromises between the speed and quality. It would be useful to have a method as a quality baseline: something that would render ideally well, even if it's very slow.
Here is such method:
With depth write enabled, and some trickery to make the front texels drawn first, this can end up in an acceptable performance.
Technically, that would be "true" Painter's algorithm, unlike the current compute scattering method.