https://hgpu.org/?p=12579
Automatic Parallelization of Tiled Stencil Loop Nests on GPUs