Beketov O. Methods of automated parallelization of cyclic operators for heterogeneous architectures of computing systems

The thesis considers the problem of automated parallelizing transformation of the nested cycles targeting a heterogeneous computational system that includes graphics processing units. The formal parallelizing method is developed. The method's advantage is that it allows to convert the amounts of data that exceeds the capacity of the on-board memory of an executing device, admits the simultaneous utilization of more than one graphics processor, and moreover, is automatable. A semi-automatic source-to-source rewriting rules based code transformation tool the aim of which is to assist in constructing a new concurrent program was implemented. To demonstrate the feasibility of the proposed approach, a series of experiments were conducted. The experiments included a set of benchmark problems tests of the developed transformation system. The tool was deployed to construct programs for solving matrix multiplication, N-body and numerical weather forecasting problems for a heterogeneous multicore cluster.