Exploring the Offload Execution Model in the Intel Xeon Phi via Matrix Inversion