For the problem of low speed of data reconciliation in the current continuous-variable quantum key distribution, this paper proposes a multi-dimensional data reconciliation scheme based on GPU and OpenCL heterogeneous computing. Moreover, we fatherly propose a static two-way cross-linked list storage hyperscale LDPC check matrix to meet the special requirements of the OpenCL′s special regulations. Experimental simulation results show that when the code length is 2×105, under the premise of ensuring effective data reconciliation in same code rate, the average decoding rate of the GPU can reach 4.2 times that of the CPU at the expense of sacrificing some precision.