I'm using Eigen to construct dynamically an optimization problem. To do this I start with empty problem description matrices and iteratively append other matrices using conservativeResize and block operations.

Accordnigly to http://eigen.tuxfamily.org/dox/classEigen_1_1PlainObjectBase.html#ac20f895ce9b73fa7447196855c0d0758 resize always reallocate data with exception of unchanged dimensions. So the proposed method for constructing matrices will reallocate memory each time matrix is appended. Is there a way to do it without reallocation? I.e. allocate enough memory for the initial matrix, so the latter append operations and thus resizing won't lead to reallocation.

You can do it manually using either:
1) a preallocated buffer, and a Map object that you remap (see the end of the doc of the Map class to see how to remap: http://eigen.tuxfamily.org/dox-devel/cl ... _1Map.html)
2) a big Matrix, and a Block<> that you dynamically reconstruct (you can use the same placement new to reconstruct the Block object)

conservativeResize was briefly discussed here

viewtopic.php?f=74&t=97410