Algebraic transformations in regular array design

Efficient parallelizing and mapping of algorithms onto regular arrays requires algebraic transformations. The paper considers the Iso-plane method for a class of algebraic transformations on the polytope model of algorithm. This method uses the partitioning of the ranges of loop indices in order to increase the dimensionality of the problem representation and a specific reordering of computations. As a result, the higher dimensional arrays with improved time complexity are produced.