Abstract: Compared to min-sum algorithm, bit-flipping algorithm for decoding of low-density parity-check (LDPC) codes can be implemented with lower complexity, but perform worse. There are increased interests on improving the performance of bit-flipping algorithm without much increase of complexity, and a recent progress is the development of the noisy gradient descent bit-flip (NGDBF) algorithm. In this paper, we propose a parallel version of NGDBF (P-NGDBF) algorithm by borrowing some ideas from parallel weighted bit-flipping (PWBF) decoding， which shows its performance favor for low row/column weight LDPC matrices. Simulations shows that the P-NGDBF algorithm has better decoding performance and its adaptive version can perform very close to the min-sum algorithm.