https://hgpu.org/?p=12515
Efficient implementation of computationally intensive algorithms on parallel computing platforms