HIGH-PERFORMANCE COMPUTING OF OIL RECOVERY PROBLEM ON A MOBILE PLATFORM USING CUDA TECHNOLOGY

This paper describes the organization of high-performance computing by using GPUs (graphics processing unit) of different devices. In this study, the following tasks are examined: the Dirichlet problem for the Poisson equation, the problem of immiscible oil dis- placement (Buckley-Leverett’s model) and mathematical model of oil displacement process by polymer-surfactant injection in a porous medium. Two algorithms have been tested: sequen- tial and parallel (GPU). The calculations are performed on a mobile device Xiaomi MiPad with NVIDIA Tegra K1 processor. The results are analyzed and compared with computations on a PC (personal computer) with a NVIDIA GeForce GTX 550Ti graphics card. The re- search output has led the authors to come to the three main conclusions: mobile devices can be used as computers to solve practical problems as well as PCs; the global memory of mobile devices prove to be ineffective in complex calculations; there is demand for practical models that use shared memory in calculations. Therefore, authors find that it is reasonably possible to conduct calculations of complex mathematical models in real-time using mobile devices. All these allow to create a mobile prototype of the being-developed ”smart” field platform for oil companies of Kazakhstan.