Parallel Algorithms for Conjugate Gradient Method of Adaptive Optics

In the wavefront reconstruction process of adaptive optics, the conventional CPU wavefront reconstruction algorithm cannot meet the real-time requirements of the system. In order to ensure that the calculation time meets the closed-loop control requirements, this paper proposes a GPU based iterative algorithm using conjugate gradient of wavefront, which is the wavefront reconstruction algorithm of CUDA architecture.By simulation using the GPU NVIDIA GeForce GTX 1070, the experimental results of adaptive optics systems with different cell numbers show that the GPU has a significant improvement on the wavefront reconstruction algorithm. When the number of elements of restoration matrix reaches 5121*5121, the algorithm improves the running speed by 76.6 times compared with the CPU algorithm, which provides a better option subsequent adaptive wavefront reconstruction.