Design Of Data Reconciliation System Based On OpenCL Heterogeneous Computing

In recent years,computer technology has developed rapidly,and information security has been increasingly valued.Quantum communication has become the focus of society's attention.Any eavesdropping behavior in the quantum communication process will disturb the quantum key.Based on the quantum unclonable theorem and Heisenberg's uncertainty principle,both parties of the communication will discover and adopt effective measures in time,thereby ensuring the unconditional security of the quantum communication system.As an important branch of quantum communication,Quantum Key Distribution(QKD)has attracted widespread attention.QKD is divided into two types,Continuous Variable Quantum Key Distribution(CVQKD)and Discrete Variable Quantum Key Distribution(DVQKD).Compared with DVQKD,CVQKD is simpler to prepare,lower in cost and less susceptible to noise.Therefore,more and more scholars at home and abroad have paid attention to it.However,compared with DVQKD,the current CVQKD has a shorter transmission distance,a lower key generation rate,and data reconciliation as the post-processing of CVQKD is of great significance to the solution of the above problems.In order to address the current low reconciliation rate of CVQKD,this paper has done the following work to accelerate the reconciliation rate:1.In order to make the parameters of the kernel functions of Open CL(Open Computing Language)conform to the Open CL specification,a static bidirectional circular linked list is designed to store large-scale sparse check matrices,and the heterogeneous parallel calculation of the Slepain-Wolf algorithm and multidimensional data reconciliation algorithm is realized.2.Implement the above two algorithms on the Open CL/GPU heterogeneous platform.Adopt NVIDIA Tesla K40C as GPU device,use Open CL and C/C ++ language for code simulation,and realize CPU/GPU data reconciliation acceleration scheme.And compare and analyze with the original CPU serial decoding rate.3.Implement the above two algorithms on the Open CL/FPGA heterogeneous platform.The DE5a-Net development board is used as the FPGA heterogeneous platform.Similarly,the GPU code in 2 is used to optimize the structure so that it can be smoothly implemented on the FPGA development board.The results are compared with the reconciliation rate of CPU/GPU and CPU,and the advantages and disadvantages are analyzed.By comparing the experimental results,it is concluded that 1)This experiment is based on the Open CL/GPU heterogeneous data reconciliation system.The SEC reconciliation algorithm can generate a raw key rate of 146kb/s,and the multidimensional data reconciliation achieves a reconciliation rate of 218.2kb/s.Compared with the pure CPU implementation,the speed is respectively increased by nearly 7.4 times and 11 times.2)Due to the lack of familiarity with FPGA pipeline task parallelism,only simple implementation of GPU heterogeneous code was implemented,which resulted in low rates of SEC and multidimensional data reconciliation system under the Open CL/FPGA heterogeneous platform in this experiment,14kb/s and 29.3kb/s,the increase of computing speed still needs to be explored by the laboratory.