Adapting of the code, simulating viscous fluids flows, to a hybrid computing systems based on CUDA-MPI

In the past few years there has been a rapidly growing interest in non-traditional architectures for high performance computing. Program development for such devices is much more complicated than for traditional general-purpose processors. Nevertheless, in the recent years, the number of users, who decided to use these technologies increased. This paper reviews the experience of a transition to a hybrid architecture for the code, solving three-dimensional Navier–Stokes equations with turbulence model. The performance measurements for the resulting program are presented. The numerical results for the flow around a NACA 0012 airfoil are presented also.