The algorithms in mldivide are not iterative, in the sense that an iterative algorithm converges to a result, improving upon it at every iteration. Instead, MLDIVIDE uses tools for matrix factorizations. For example, depending on the specific matrix, you might find a Cholesky, a QR, an LU, etc. Even in the case of sparse matrices, the algorithms in mldivide are still not iterative.
Some other tools, such as LSQR, are indeed iterative. But LSQR is not part of mldivide.
