
20. Efficient Layered Fragment Buffer Techniques 289
Figure 20.10. Transparent layered grids of polygons.
polygons (500–1000) approximately constant. Rendering times are similar for both
linearized and linked-list LFBs, as sorting is a common operation. The sorting time
becomes dominant after approximately 50 layers. Simply declaring and populating
the sorting array (no sorting) with 256 vec4 elements causes a 3–4× slowdown,
compared to blending unsorted fragments directly from video memory (no local
array). As expected, O(n
2
) insertion sor t is faster for small n, for example, in the
dragon and atrium scenes. We expect most scenes to have similar