https://hgpu.org/?p=12097
targetDP: an Abstraction of Lattice Based Parallelism with Portable Performance