Studying the Efficiency of the Parallel Algorithm for Solving the Eigenvalue Problem Implemented in the LUCKY-A Computer Code

Аннотация

The efficiency of a parallel algorithm for solving the eigenvalue problem implemented in the LUCKY-A computer code is studied. The parallel algorithm is developed for application on supercomputers using the MPI standard of data exchange between parallel processes. The test problem–model of a fast breeder–is studied with the use of the library of 21-group cross sections. Results of solving this problem obtained by the LUCKY-A, TORT, and MMKKENO computer codes are presented. The efficiency of the parallel algorithm is determined as a function of the number of space subdomains in the computational domain.