@rasterize Right: if you just use a single set of equally-sized tiles, then you get something like a uniform grid: a grid of larger boxes, each of which contain several primitives (in this case, individual fragments). You could of course take it up a notch by adding an additional level of even bigger boxes on top of this one, which starts to look more like a quadtree. Modern GPUs, AFAIK, use only a single "level", especially for things like "early Z-cull", i.e., doing a coarse-level depth check before going on to rasterize fragments at a fine level.
Uniform grid!
@rasterize Right: if you just use a single set of equally-sized tiles, then you get something like a uniform grid: a grid of larger boxes, each of which contain several primitives (in this case, individual fragments). You could of course take it up a notch by adding an additional level of even bigger boxes on top of this one, which starts to look more like a quadtree. Modern GPUs, AFAIK, use only a single "level", especially for things like "early Z-cull", i.e., doing a coarse-level depth check before going on to rasterize fragments at a fine level.