Parallel realization of statistical simulation and random number generators
Михаил Александрович Марченко, Guennady A. Mikhailov · Russian Journal of Numerical Analysis and Mathematical Modelling · 2002
- In this paper we consider various aspects of parallel realization of the Monte Carlo method algorithms on multiprocessor computation systems. In particular, we consider the distribution of ‘statistical tests’ among various processors and the correlation of the corresponding results with the aim to effectively estimate functional relationships in the metric C . We developed the modification of an ‘astronomically’ long-period ‘congruential’ program pseudorandom number generator. It allows us to effectively distribute generated numbers essentially among an arbitrary number of processors. We conducted multidimensional uniformity tests for the modified generator by the ‘c-square’ criterion to the seventh dimension. All the tests gave satisfactory results. The corresponding computational routines in Fortran 90 are given in the Section 4.