https://hgpu.org/?p=6408
Overlapping Computation and Communication for Advection on Hybrid Parallel Computers