We present a new method to simulate repetitive ferromagnetic structures. This macro geometry approach combines treatment of short-range interactions (i.e. the exchange field) as for periodic boundary conditions with a specification of the arrangement of copies of the primary simulation cell in order to correctly include effects of the demagnetizing field. This method (i) solves a consistency problem that prevents the naive application of 3d periodic boundary conditions in micromagnetism and (ii) is well suited for the efficient simulation of repetitive systems of any size.Introduction One often wants to quantitatively understand the behaviour of highly regular extended periodic magnetic structures created, for example, by selfassembly methods. As these structures often are too large for full micromagnetic simulation, some physically justifiable assumptions have to be made to reduce the computational complexity.One attractive simplification is to define a 'primary' simulation cell C 0 in the bulk of the sample (which may contain one or multiple elementary lattice cells) and demand that the magnetization in every other 'image' cell C j is constrained to behaving as a translated copy of the magnetization in cell C 0 . While this approximation loses long range multicell structures (such as the domain wall structure of extended materials) and field deviations near surfaces, it is useful in other situations such as magnetic films in strong external fields. This idea is usually described by the term 'periodic boundary conditions' (PBC). We note that there is a conceptual difference between using PBC to remove surface artefacts in multi-particle statistical mechanics simulations (e.g. of a gas of hard spheres) and using PBC to describe real periodic structures with long-range interactions (such as magnetic repulsion).While no problem arises for structures that are periodic in one or two directions only, micromagnetism requires a modification of the PBC approach for 3d periodic systems, as taking the infinite size limit leaves essential information about the shape of the (now 2d) surface unspecified which here matters for long-range interactions (see also discussion in section 3 in [1]).To show why the shape matters, we consider two different structures s A and s B made of one billion individual cells each: s A being a cube of 1000×1000×1000 cells, and s B being a film of 10000 × 10000 × 10 cells. If both structures are homogeneously magnetized, a cell in the center of each experiences a demagnetizing field H = −T M . The demagnetizing tensor T must satisfy (a consequence of the Maxwell equation div B = 0): tr T = 1. For s A , we have T (A) = diag(1/3, 1/3, 1/3), while for s B , we get T (B) = diag(0, 0, 1). The form of T depends on the shape