https://hgpu.org/?p=16564
Acceleration of Block-Aware Matrix Factorization on Heterogeneous Platforms