The mixing of non-spherical particles is often encountered in various industrial fields, but its mechanism is not well understood. Because of the complexity of contact detection for non-spherical particles, most simulations on particle mixing using the discrete-element method (DEM) are restricted to spherical particles. In this work, an efficient algorithm is used for the contact detection and the description of contact forces, with which the flow and mixing of geometrically exact sphero-cylinders in rotating drums was simulated in parallel computing using graphics processing units (GPUs). The specific shapes of the sphero-cylinders affect their dynamic behavior in mono- and bi-disperse systems, including the dynamic angle of repose, mixing constant, and segregation pattern. Certain kinds of preferential particle orientations exist, which are affected by the rotational speed and particle shape. The density difference leads to segregated cores of a higher density in the granular beds, whereas the effect of particle shape is secondary in this process. The joint effect of shape and density on particle distribution in the steady state can be used to improve the homogeneity of bi-disperse sphero-cylinder mixtures that differ in shape and density. (C) 2018 Elsevier B.V. All rights reserved.