GPU utilization for speed up the solution of the three-dimensional engineering tasks of open flows hydraulics

The work is dedicated to adaptation for the graphic accelerator (GPU) of the numerical open flow simulation algorithms in the planned and 3D setting. Algorithms are suitable for the simulation of heat or impurities transport in the reservoirs, for the calculation of the wind waves transformation and other tasks that require a long time, even when parallelizing on the CPU. A special feature is the GPU implementation of the algorithms based on the multi-layer control volume method, taking into account the real depth pressure profile. Effectively implemented on the GPU simple iterative refinement method of the pressure profile along the vertical is used for 3D-tasks. Some of the OpenCL language peculiarities, increasing the computing efficiency on the graphics accelerator are described. The calculations examples with comparison the elapsed time for CPU and GPU are presented.