- abstraction layer for selecting CUDA block and grid sizes
  - line based (was implemented before)
  - block based (new, more flexible)
-  new conditional (if/else) ast node, which is necessary for indexing schemes (guarding if)