As the resolution of the computational spatial light modulator has become larger, the computational complexity of the three-dimensional dynamic display of the hologram is also higher and higher, which makes new requirements for the holographic calculation speed. The fast calculation of hologram by slice-based method is realized by means of the graphics processing unit (GPU) parallel computation processing method, which accelerates the calculation of the Fresnel diffraction transform algorithm by taking advantage of the GPU parallel multi-threading and the image two-dimensional Fourier transform in tomography. At the same time, by calling the GPU underlying resources and the stream processing of the program in compute unified device architecture (CUDA), the intermediate delay wait is effectively reduced. The comparison of the calculation speed shows that the calculation speed is greatly improved compared with the calculation on the CPU, and the method based on GPU parallel calculation is about 10 times faster than the method based on CPU calculation.