Parallel execution of program loops (Q796290)

The appearance of parallel computing systems necessitated the development of automatic paralleling of regular sequential program loops. Efficient transformation of such loops offers significant saving of program execution time. Here we propose a linear transformation method which allows paralleling of a much more extensive class of loops than other known methods. By applying this method, the space of iterations for which the original loop body is to be executed is partitioned into parallel subspaces of any dimensionality for all of whose points the loop body can be simultaneously executed.