the Intel MKL DGEMM function for matrix-matrix multiplication is highly tuned for small matrices. To eliminate overhead, Intel MKL provides a compiler flag to guarantee that the fastest code path is ...