ok so i want the fast stuff, and the count at the same time.
thats why i allocated my 480,000,000 vertices in 12 calls to get em all out, didnt bother projecting, cause i like 2d stuff.
Just think if I did 16 texture reads in each vertex, god it would be a 32th of the pace, which means if im doing something faster with a gpu reading textures going 32 times faster, that means im going 1024x speedup using vertex stride instead of texture reads.
something to be proud of, hitting half a billion paramaters, in full 60hz realtime. im actually just making an accelleration structure for a neural net, not like it hasnt been done before.