For reconciliation of Gaussian quantum key distribution, optimal quantization intervals of continuous variables are searched to maximize the mutual information between Alice and Bob. Based on both sliced error correction (SEC) and multilevel coding/multistage decoding (MLC/MSD) protocols, low density parity check (LDPC) is employed in each level of coding streams. A one-time multistage iterative information update formula for MSD algorithm is also derived. In the implementation, double cross-linked list is used to store sparse matrix H of LDPC. C language is also used to realize the whole reconciliation process. These greatly reduce space complexity and speed up reconciliation process. Simulation results show that the proposed algorithm can reconcile 2×105 continuous quantum variables reliably when signal-to-noise ratio of receiver is above 4.9 dB, with reconciliation efficiency of 91.71%. On a server with 2.4 GHz CPU and 32 G memory, the speed of the reconciliation reaches 7262 bit/s.