Operator library for three-dimensional discrete schemes of mathematical physics problems using GPUs with CUDA architecture

This paper describes a library of operators to work with grid functions, aimed to simplify the calculations on three-dimensional index grid. The library is designed so that the details of its implementation are hidden from the user, allowing effective implementation on machines with different architectures (parallel, hybrid, etc.). The library creation raises several goals: the approximation of the appearance of the program to the theoretical formulas, the relative ease of use, including the work on the graphics card with CUDA architecture, computational efficiency.