Calculation performance improvement of particle dynamics on parallel systems

In the article, the source code modifications of software COSY Infinity for calculation performance improvement on parallel systems with shared memory are described. These modifications are made manually without using techniques of automatic and automated parallelization. For multiprocessor systems with shared memory OpenMP technology is used. For computing on graphics processors NVidia CUDA is used. Also, in the article hybrid versions with a bunch of technologies OpenMP and MPI used for calculations on cluster systems and using of OpenMP and NVidia CUDA for personal computers are presented. Evaluation of the effectiveness of the obtained versions is made by comparing the time of the calculation of the beam dynamics in an electrostatic ring including the spin. Testing has shown an increase in computing performance using OpenMP from 5 to 10%. When using simultaneously OpenMP and NVidia CUDA performance increase reached 75%. Bibliogr. 9. Il. 4.