Abstract:
The paper is dedicated to the problem of mapping of computational algorithms onto distributed memory multiprocessor computers. A procedure of parallelization is presented. The procedure uses new more general conditions of data localization that take into consideration more possibilities of effective use of data in memory.