Abstract:
Operator programming method allows the compact writing of mathematical formulas in programs and helps to transfer the programs to parallel architectures, such as NVidia CUDA and Intel Xeon Phi. Earlier the operator programming method was implemented for regular three-dimensional Cartesian grids and tree-dimensional locally adaptive grids. In this work, this method is transferred to three-dimensional tetrahedron meshes. This example illustrates the possibility of method implementation on arbitrary tree-dimensional meshes. Besides, in this work we demonstrate the usage of template metaprogramming methods of the C++ programming language to speed-up calculations.