Modeling of electron transport in matter with the use of hybrid supercomputers

Distributions of electron characteristics for modeling their transport in matter are obtained with the use of cross-section data on elastic and inelastic processes. The developed model does not use the well-known slow-down approximation and the multiscattering theory of electron transport. Statistical algorithms for modeling of electron transport are proposed to perform numerical experiments on hybrid supercomputers. In particular, these algorithms use the NVIDIA CUDA technology. A number of numerical experiments are performed to study the interaction between electrons and a target matter of an X-ray tube on an MVS cluster (http://www.kiam.ru/MVS/resourses/mvse.html) and on a new supercomputer K-100.