With the growing demand of large-scale aerospace equipment, efficient and high-precision large-scale topology optimization design becomes focal point in this field. To address the issues of massive computational load and low efficiency in existing large-scale topology optimization design, study of full-process accelerating design of topology optimization based on GPU parallel computation is carried out. The proposed method accelerates the entire process including meshing, stiffness matrix computation and assembly, and solving of finite element linear systems, achieving efficient and precise voxel meshing and high-efficiency solutions to finite element process. Moreover, the proposed method parallel accelerates the sensitivity filtering process aiming at the accelerating demand by the topology-optimization design process. The design case of a 3-million-voxel attitude thruster model demonstrates that compared with the parallel speed-up ratio of topology optimization by the Abaqus 2022 software, that of the proposed method increased 1259%, and the similarities of the two methods are extremely high, verifying the validity and practicality of the proposed method.