Current continuous-variable quantum key distribution systems suffer from low computing speed for data reconciliation. Herein, we address this problem by implementing a parallel acceleration for a multidimensional data reconciliation algorithm based on a central processing unit/graphics processing unit (CPU/GPU) heterogeneous platform. To meet the special requirements of heterogeneous parallel computing, we propose a static two-way crosslinked list to store a hyperscale low-density parity-check matrix. We also propose a parallel reconciliation algorithm. A simulation experiment is carried out on the heterogenous platform with a code length of 2.048×10 5. Reconciliation speed, key transmission distance, and reconciliation efficiency are calculated based on the simulation results of the convergence signal-to-noise ratio and time of reconciliation. Results show that when the code length is 2.048×10 5, the reconciliation speed of parallel acceleration on the CPU/GPU heterogeneous platform is five times faster than that on the CPU platform.