A Hybrid PSO with Load-Balanced SPV for Real-time Task Scheduling on Chip Multiprocessor
Song Chai · Journal of Information and Computational Science · 2013
In this paper, a hybrid PSO heuristic and an improved solution representation rules are proposed for the real-time task schedule problem on Chip Multiprocessor (CMP). The hybrid PSO incorporates a Neighbor Search process at the final stage of each move in a standard PSO, and highly improves the result of optimization. The new solution representation rule called the Load-Balanced SPV, evenly assigns tasks to the chip multi-processors according to their total loads, and produces more balanced schedule than the original SPV rule. A comparative performance of our proposal is studied. Through a series of simulations, we demonstrate that our proposal produces much better balanced schedule as well as better overall fitness that the original ones.