update documentation

src/main/java/me/jellysquid/mods/sodium/client/render/chunk/gfni/ComplexSorting.java

@@ -9,6 +9,15 @@
 import me.jellysquid.mods.sodium.client.util.collections.BitArray;
 import me.jellysquid.mods.sodium.client.util.sorting.MergeSort;
 
+/**
+ * This class contains the sorting algorithms that do topological or distance
+ * sorting. The algorithms are combined in this class to separate them from the
+ * general management code in other classes.
+ * 
+ * Rough attempts at underapproximation of the visibility graph where the
+ * conditions for visibility between quads are made more strict did not yield
+ * significantly more robust topo sorting.
+ */
 public class ComplexSorting {
     private ComplexSorting() {
     }

src/main/java/me/jellysquid/mods/sodium/client/render/chunk/gfni/GFNITriggers.java

@@ -10,6 +10,8 @@
 /**
  * Performs Global Face Normal Indexing-based triggering as described in
  * https://hackmd.io/@douira100/sodium-sl-gfni
+ * 
+ * Distances are stored as doubles and normals are stored as float vectors.
  */
 class GFNITriggers implements SectionTriggers {
 	/**

src/main/java/me/jellysquid/mods/sodium/client/render/chunk/gfni/TranslucentSorting.java

@@ -13,8 +13,6 @@
  * of translucent data objects for each sort type and delegates triggering of
  * sections for dynamic sorting to the trigger components.
  * 
- * Distances are stored as doubles and normals are stored as float vectors.
- * 
  * TODO:
  * - many sections can be marked as needing an update but they are only actually
  * scheduled for sorting when the RenderSectionManager makes them visible. This
@@ -45,10 +43,6 @@
  * - re-use distance sort data to accelerate future distance sorting. Sorting
  * mostly sorted data is probably faster than sorting unsorted data (or whatever
  * data comes out of the mesh).
- * - investigate why a small epsilon of 0.001 makes many more sections sortable
- * when it's added as a "slack" distance to the complex sorting's quad
- * visibility check. This suggests that the quad visibility check or the
- * quantization of vertex positions is not working correctly.
  * 
  * @author douira
  */