Computational efficiency of GPU-based uncompressible flow simulation

The paper considers increasing of computational efficiency by means of graphics processors. The numerical method is based on pressure-velocity coupling by pressure correction equation. CUDA is used for calculation on graphical processors. Parallel calculations on graphics processors, central processors or both graphical and central processors are used with computational domain decomposition. Steady laminar flow in a container with rotating end wall and unsteady laminar flow around a circular cylinder are used as tests. Higher computational efficiency of calculations on graphics processors as compared with calculations on central processors is shown.