In adaptive optics system, wavefront computation refers to a class of processing consisting of wavefront slope computation, wavefront reconstruction and wavefront control algorithm. We studied the realization of the optimization mapping of the wavefront computing algorithm from the representive layer to the architecture layer. Two SIMD-based parallel algorithms, the doubling algorithm and the partitioning algorithm are proposed according to the existing parallel computing theory. On the basis above, the MAC cycle and I/O cycle are used to scale the performance of the two algorithms. Their computing efficiencies on the SIMD array processor are analyzed and compared in detail.