https://hgpu.org/?p=11218
Toward a Generic Hybrid CPU-GPU Parallelization of Divide-and-Conquer Algorithms