In this paper the midpoints of implementation of arbitrary-precision arithmetic library with using CUDA technology for cryptographic applications are considered. The ingenious approaches to solving problems such as limited speed of memory access, complexity of parallel implementation of algorithms for basic arithmetic operations and synchronization between the blocks of parallel threads are proposed. Tests to compare the effectiveness of simple arithmetic operations (vector addition and calculation the arithmetic progression sum) on the CPU and GPU were conducted. The nuances of implementation the adding operation for long integer numbers using GPU are examined.
arbitrary-precision arithmetic, GPU, CUDA, parallel algorithms, a vector computer, graphically accelerator