Coherent De-dispersion Algorithm for Pulsar GPU Based on Zero-copy

The radio pulse signal is affected by the interstellar medium during the transmission process, which will cause the profile to be broadened and deformed. In the research process, the signal needs to be de-dispersed. This paper designs and implements a zero-copy-based GPU coherent de-dispersion algorithm for pulsar data, uses device memory mapping to eliminate the copy cost from host to device, and uses CUDA's cuFFT library to perform multi-BATCH Fourier transform to improve DFT efficiency. Multithreading is used to realize the accelerated calculation of the transfer function. The experimental results show that, compared with traditional CPU and GPU algorithms, the algorithm proposed in this paper performs well in large amounts of data.