https://hgpu.org/?p=6593
Simultaneous Branch and Warp Interweaving for Sustained GPU Performance