Fast and accurate calculation of the diffraction field is a challenging problem in computer-generated holography. In this work, we present the performances of three fast algorithms that can be used in accurate computation of the diffraction field from a point cloud object on a display device which has pixelated structure. Performances of the algorithms are evaluated according to computation speed and error on the reconstructed object. First algorithm is based on scaling of a pre-computed one-dimensional (1D) kernel, the second one utilizes a look-up-table (LUT) which is formed by 496 pre-computed 1D kernels, and third one is a hybrid algorithm which uses LUT of the second algorithm with the scaling function employed in the first algorithm. 1D kernels are used to keep the allocated memory space by the LUT in manageable sizes. Third algorithm yields the best error performance compared to other two, but its computation time performance is not as good as the others. Even if, the second algorithm excels the first and the third ones in computation time, it gives the worst error performance among the presented algorithms. The computation times of all three algorithms are improved by having parallel computing.