A parallel simulation method for underwater wireless optical channel simulation is proposed based on open multi-processing (Open MP) and compute unified device architecture (CUDA) to address the computational complexity and low computational efficiency of the Monte Carlo-based model. The simulation calculation efficiency is improved by transplanting computationally intensive parts to threaded parallel computing. Three optimization schemes are introduced to accelerate merging by eliminating invalid photons, limiting high-scattering events, and reducing the amount of data exchange between main and video memories, further improving the simulation efficiency. The acceleration effects in various water types, computing environments, and photon numbers and distances are evaluated. The parallel calculation efficiency of graphics processing units has an acceleration effect of up to 300 times, whereas that of central processing units has an acceleration effect of 90 times compared with traditional serial simulation.