Design Of LDPC Decoder Based On GPU For Reconciliation

Quantum Key Distribution utilizes "The Quantum No-cloning Theorem" and "The Uncertainty Principle" to ensure that the information is absolutely safe.It has become the important option of quantum secure communication and the future development direction,has been attach great importance at home and abroad.Reconciliation is an essential part of quantum key distribution.On the quantum channel,reconciliation protocol is used to remove and correct errors due to the thermal noise of the channel and the presence of eavesdroppers in the original information sequence codeword.Reconciliation can also be seen as asymmetric distributed source coding,part of the information compression combined with another part of the information decoding.In the reconciliation protocol for continuous variable quantum key distribution system,component code commonly used performance is good,close to Shannon limit,code structure flexible LDPC.For the low computing speed and long latency of reconciliation for current continuous variable quantum key distribution,this paper designed CPU&GPU-parallel reconciliation algorithms based on the SEC protocol to speed up decoding computing.In order to raise decoding speed without sacrifice reconciliation efficiency,we employed a static two-way cross linked list to efficiently store large scale sparse parity matrix?The experimental simulation results show that 2x105 continuous variable quantum sequences are reliably reconciliated when signal-to-noise ratio of communication channel is over 4.9 d B,with reconciliation efficiency of 91.71%.On a platform with Geforce GT 650 M GPU,2.5 GHz CPU and 8G memory,the speed of the decoding rate reaches 16.4kbit/s.Relative to the only CPU platform,computing speed increased by more than 15 times.