The stochastic parallel gradient descent (SPGD) algorithm can be used to correct phase distortion by optimizing the system performance index directly, so it is a promising control algorithm for adaptive optics (AO) system. In the point-source adaptive optics based on SPGD algorithm, the quadratic sum of intensity, the mean radius and encircled energy are usually used as the system performance index. The relation between the three performance index and the RMS value of aberrant wavefront are analyzed using numerical simulation. Then an adaptive optics experimental system is set up and the static phase-distortion correction experimental is carried out to research three performance index of SPGD algorithm. The experiment results show that the mean radius is the most efficient as the system performance index compared with the other two, which is accordant with the result of numerical simulation previously. The relation between the speed of convergence and control channel number is also researched. The experimental results show that the iteration amount required by convergence of performance index curve is increased remarkably when the control channel number is increased, and it is approximately proportional to square root of control channel number.