https://hgpu.org/?p=18116
High-performance sparse matrix-matrix products on Intel KNL and multicore architectures