With the arrival of parallel computing era, parallel computing of formal concepts has become a hot issue in the field of formal concept analysis. This paper proposes a parallel concept computing algorithm by means of the graph characteristics of an attribute topology used in representing formal context. First, according to the parent relations, the bottom-up decomposition of attribute topology is conducted to generate sub-topologies. Then, concept-couplings among sub-topologies are removed based on the correlations in attribute-pairs in order to ensure the independence of the sub-topologies when carrying out concept computing and then to avoid large time consumption of the merging operation in later stage. Finally, all the concepts without repetition can be calculated by accumulating directly all the concept-sets computed in different sub-topologies. The experiment shows that the approach proposed in this paper can not only obtain all the concepts without repetition, but also improve the computational efficiency in accordance with the hardware platform and reduce the time required for the concept calculation.