For sparse matrix operations that is the most common pattern of access though. For linear algebraic operations on dense matrices, BLAS and LAPACK takes care of most of the issues and is super fast. I think the times I do not use BLAS/LAPACK (directly or indirectly) are when I have to apply operations on matrices that are not the typical linear algebraic operations. This comes up fairly often too in my line of work which is machine learning.