Yu. A. Shichkina

Ontological modeling for building parallel programs based on data is offered as application. Parallezing based on data depends on a data model, and relational model in that case lets to increase quality of parallezing due to realization of definitive restriction types. Such restrictions are resulted from analysis of data domain modeled by an information system. Data dependencies are one of the tools of information formalization retrieved as the result of such analysis. Transformation algorithm for optional selection of similar objects - groups and links between those groups is defined. Possibility of Ontological model compliance with matrix representation is shown. Matrices depicting ontological modeling will be sparse in that case. Bearing in mind that tasks, based on matrix calculation, can be most easily parallelized, parallel analogues and optimization methods can be applied to the suggested algorithm