https://hgpu.org/?p=1065
Program optimization space pruning for a multithreaded gpu