https://hgpu.org/?p=12178
A Performance Model for the Communication in Fast Multipole Methods on HPC Platforms