Adapting Matrix-Vector-Multiply Wavefront Reconstruction to SMP

In adaptive optics systems, the process time of wave front controller is a critical parameter. For a fixed frame rate, decreasing the process time results in higher control bandwidth. Among the process steps of a wave front controller, wave front reconstruction is a time consuming task. In this paper, we parallelized wave front reconstruction algorithm on a cutting edge shared-memory eight-core DSP to reduce process time. The conducted experiment shows that even using the most computation and memory intensive matrix-vector-multiply (MVM) reconstruction algorithm, performance requirements can be met with reconstruction latency around 130us in a 913 actuators AO system. With parallelization of the reconstruction algorithm, it's possible. to apply high frame rate in the system and easy to scale up to larger systems with more actuators.