摘要：For complex technical objects with a large number of input variables computing time to build a mathematical model quite noticeable. Analysis method of construction of empirical models of optimal complexity based on genetic algorithms showed that this algorithm has an internal parallelism, which allows the development of an effective program implementation, which will reduce the cost of computer time. The most expensive operation of the algorithm is to solve a system of linear algebraic equations. These operations are performed repeatedly, so to reduce computer time parallel algorithms have been developed for their implementation. Existing algorithms in the evaluation of their performance take into account only the time used for solving a system of linear algebraic equations and do not include the transmission or reading data, which can be much larger, so such estimate can not be an adequate criterion of performance of the algorithm. We analyzed the performance of existing algorithms for parallel solution of linear algebraic equations with taking into account time to transfer and read data. Implemented the algorithm, which allows to significantly reduce the cost of computer time to solve systems of linear algebraic equations.
关键词：эмпирическая модель; параллелизм; C ++ MPI; C #; NET; TPL.
