https://hgpu.org/?p=8679
Sparse Matrix Multiplication using CUDA and Mex Interface