https://hgpu.org/?p=13158
A CUDA implementation of the High Performance Conjugate Gradient benchmark