Solving the nonlinear Schrödinger equation, which describes the signal evolution in an optical fiber, can be a time-consuming task. This holds in particular for long test signals, which have to be used to account for the channel memory in high-speed communication systems. Hence, the optimization of optical links, which usually requires many evaluations of the signal propagation, can take very long. In this article we describe how the simulation of signal propagation can be significantly accelerated by using commodity graphics hardware. We have carried out the simulation of a benchmark scenario with and without the aid of the graphics processing unit (GPU). The comparison has shown a speed-up of simulations by factors up to $80$ if computations are performed on the GPU rather than on a state-of-the-art CPU.     «

Solving the nonlinear Schrödinger equation, which describes the signal evolution in an optical fiber, can be a time-consuming task. This holds in particular for long test signals, which have to be used to account for the channel memory in high-speed communication systems. Hence, the optimization of optical links, which usually requires many evaluations of the signal propagation, can take very long. In this article we describe how the simulation of signal propagation can be significantly accelera...     »