Using an Evolutionary Algorithm for Scheduling of Two-Level Nested Loops

The need for high computational speed and power in a majority of scientific applications fuels the incentives for gaining the computational power of several processors to raise the execution speed of programs. Furthermore, the presence of sequential programs, once very costly generated, provokes the engagement of tools known as "super-compilers" for automatic conversion of sequential codes into parallel codes. Super compilers can detect the hidden parallelism in programs and next convert a sequential program into a parallel one. Most of computer programs use nested loops. Parallel execution of associated loops might accelerate the execution of in-question programs. So the parallelization of nested loops is a key challenge in shortening the computer program run-times. One of underlying stages in parallelization is scheduling tiled space for iterating nested loops. As the problem is a NP-Hard one, using traditional search methods for solving such programs does not fit the case. So evolutionary algorithms must be engaged to solve these kinds of problems. In this paper, based on the general wave-front method, one distinctive approach is developed. Practical results show that our solution approach which is inspired by heuristics provides better solutions than previous approach alternative solutions in the literature.