Research and development of a basis for the computation platform and application programming interface (API) for automated numerical simulation of large scale aerodynamic and hydrodynamic problems on petaflops supercomputers.
The project was aimed at the creation of an experimental platform for numerical simulation on the top of the OpenFOAM library for heterogeneous computer systems with graphical processing units transferring the most resource-intensive computations to the graphical processing unit using CUDA technology and managing central processing unit and graphical processing unit interaction.
The key feature of the project is the fact that the created parallel version was tested on a range of model problems and large scale aerohydrodynamics problems with the use of clusters and graphical processing units.
The developed experimental sample of the platform allows numerical simulation of:
- sub-, trans-, super-, and hypersonic flows via cut-through computing approach;
- turbulent energy generation and dissipation;
- multiphase flow of continuous and disperse medium within multi-rate approach;
- immiscible fluids flow with contact surfaces adjusted for phase interaction;
- flows with physicochemical transformations, including combustion;
- movements of solid and deform bodies in computation domain;
- fluid and construction interaction within bilateral junction with external programs of deflected body mode calculation.
Furthermore, the developed software provides an application programming interface (API) to support programming with the use of GPU accelerators and allows performing computations on meshes up to 1 billion cells.
The project was implemented under the Program "Research and development in priority areas of Russia scientific and technological complex development for 2007-2013".