Parallelism in Matrix Computations