The key issue of atmosphere turbulence simulation is how to generate multi-layer turbulence phase screen with high precision. Classical spectral inversion method needs to take up lots of computer resources.In this case, it leads to impossible parallel computing and slow calculation.In order to remedy these drawbacks,a new method was proposed and implemented based on parallel computation and multi-disk memory mapping.The basic idea of the method was that by using multi-disk to temporary storage large files,many data blocks were mapped into memory for parallel computing.The function interfaces provided by operating system were used to implement block automated memory mapping mechanism without human intervention.Experiments showed that both the accuracy of simulation experiment and the speed of generation of turbulence phase screen have greatly ascended.Moreover,the method could not only be used to generate big data turbulence phase screen parallel,but also could be applied to general big data in scientific computation and computer simulation.
