There are many types of harmonic analysis algorithms for grid signals, which are suitable for different load scenarios. However, their performance are different, There is still no unified performance evaluation index system, which makes many users unable to choose the most suitable harmonic analysis algorithm according to their specific needs more reasonably and quickly. In view of this, this paper compares the computational performance of 621 kinds of harmonic analysis algorithms for grid signals. Specifically, many window functions are classified according to different characteristics in frequency domain. According to the principle, the harmonic analysis algorithms for grid signals are divided into three categories: windowed FFT interpolation, windowed FFT and harmonic grouping algorithms. Furthermore, the computational accuracy as well as the computational cost of hardware and software of these algorithms are compared and analyzed. The paper also investigates the effects of nonsynchronous sample and noise on the computational accuracy of the algorithms. Based on the analysis results of typical signals, namely voltage of the power supply system, the electric arc furnace current and the wind turbine current, recommended algorithms are recommended. Finally, the computational performance of the steady-state harmonic analysis algorithms on the measured signals with typical dynamic changes is tested. The conclusions obtained in this paper provide convincing guidance and help for different users to select the appropriate harmonic analysis algorithms and to carry out related research according to their specific needs.