Computer simulation of optical wave propagation with the use of parallel programming

Methods and peculiarities of parallel algorithms for numerical simulation of optical wave propagation are considered. A scalar parabolic equation for the complex amplitude of monochromatic-wave field was solved numerically using the Fourier transform method for homogeneous media and split-step Fourier method for inhomogeneous media. Two parallel algorithms have been constructed—with the use of OpenMP technology with the MKL library for Intel multicore processors and CUDA technology for NVIDIA graphics accelerators. Speed comparison of these algorithms with each other and with a conventional sequential two-dimensional algorithm from the FFTW library is carried out by calculating the average number of test task solutions per second. It is shown that the parallel algorithms have a significant speed advantage (by tens of times) over the common sequential algorithm; and the larger the grids in a computation task, the more significant the advantage. Comparison of the above parallel algorithms shows the following: the approach based on the OpenMP technology holds the lead for grids of up to 1024 × 1024 in size, while the approach using CUDA technology was faster for large grids (from 1024 × 1024 or larger). The results are discussed, and recommendations on switching from sequential algorithms to the parallel ones are given.