Increasing the energy efficiency of a computational system by using automatic software optimization

Power consumption of computational systems is caused by the different sources. Dynamic power dissipation occurs when digital signal changes its level. Optimized software minimizes logic levels switching. Software can be optimized in the different ways. Loop transformations are powerful range of the approaches of the optimization methods. Subject of the paper is tiling method of the loop transformation. The results of the optimization by tiling method and parallelization for the test applications are represented. Both time of the processing and energy consumption are measured and analyzed.