A salient feature of additive manufacturing is that the cost of fabrication, to a large extent, is independent of geometric complexity. This opens new opportunities for custom-designing parts both at a macro and micro-level. An elegant and powerful method of designing custom-parts is through topology optimization.
While the theory of topology optimization is well understood, current methods can be extraordinarily expensive. The focus of this paper is on efficient microstructural topology optimization for 3d-printing. In particular, the computational bottle-necks in microstructural topology optimization are identified. Then, a framework that not only eliminates these bottle-necks, but incorporates other significant improvements, is developed. The framework is demonstrated through numerical experiments involving microstructures with millions of degrees of freedom, using multi-core CPUs and NVidia GPU.