https://hgpu.org/?p=14243
Model-based optimization of MPDATA on Intel Xeon Phi through load imbalancing