Heterogeneous Assignment of Functional Units with Gaussian Execution Time on A Tree

Time and cost are two critical performance metrics for computing systems; notwithstanding, they generally associatively change in a reverse direction. In a real-world scenario, the execution time may not be fixed due to various influencing factors. Gaussian distribution is an effective way for many cases to model the execution time by random variables. In this paper, we investigate the method of minimizing the total cost while satisfying timing constraints for heterogeneous systems with Gaussian distributed execution time on a tree, which is called HAP-G-T (Heterogeneous Assignment with Probability - Gaussian - Tree). The core algorithm in our approach is a highly efficient heuristic algorithm. Our investigation also implements experiments to evaluate the effectiveness of our approach. The experiment results depict that our solution can significantly reduce the total cost for the tree case compared with the hard real-time.