https://hgpu.org/?p=5477
Parallel implementation of conjugate gradient method on graphics processors