https://hgpu.org/?p=7911
Implementing the Approximate Message Passing (AMP) Algorithm on a GPU