I’ve made everything pooled (Faces, Caches, Vec3s).
I’ve modularised the entire mod.
Rendering is now completely split from mesh generation.
I build my caches as meshSize + 1
on all axis, meshSize = 18
for Surface Nets and meshSize = 17
for all other mesh generators.
I now use 1 PooledMutableBlockPos
for my entire mesh gen & rendering instead of more than 500,000 BlockPos
s.
This means I do 19*19*19 (6859) lookups with 1 pos for Surface Nets and 18*18*18 (5832) lookups with 1 pos for all other mesh generators.
This makes my current approach more than 61x faster for Surface Nets and 72x faster for all other all other mesh generators†.
All transparency issues are now solved!
†This does not include the performance gains I get from only using 1 PooledMutableBlockPos
instead of many, many BlockPos
s.