In my parallel programming class we used several techniques to increase the speed of matrix multiplication, and compared them. I vaguely remember using OpenMP and cuda. I need to look into my backups to see if I still have those codes. Specially the cuda one, I wonder how similar it is to tensors