https://hgpu.org/?p=11921
Sparser, Better, Faster GPU Parsing