Parallel Implementations of the Cholesky Decomposition on CPUs and GPUs