African Vulture with Aquila Optimizer AVAO for high performance energy efficient task allocation and DB RNN based optimum core prediction for multicore systems

Abstract

Reduced power consumption and improved performance have become newlinesignificant metrics in the development of multicore processors. Due to the ceasing newlineof Moore s law and Dennard scaling, reducing power budget without newlinecompromising the overall performance is considered as a predominant limiting newlinefactor in the design of multicore systems. Modern computing system must newlineconsider energy efficiency due to increasing power demands and environmental newlineconcerns. Rapid increase in mobile embedded systems necessitates a more newlinecomprehensive optimization of power-saving capabilities. newlineThis thesis proposes an energy aware hybrid meta-heuristic algorithm newline African Vulture with Aquila optimizer (AVAO), a combination of African newlineVulture Optimization Algorithm (AVOA) and Aquila Optimization Algorithm newline(AOA) which shows great potential in optimizing energy consumption and newlinemakespan in multicore environments. The suggested model is executed via newlineMATLAB; it addresses energy consumption, temperature deviations and newlinemakespan of multicore systems. An added feature of the proposed method is that newlinethe algorithm considers capacity of cores and intensity of workloads, so that more newlinecomputationally intensive tasks can be assigned to cores having greater capacity. newlineIt will reduce the amount of dark silicon because all the available cores are newlinerunning simultaneously. The simulation results show that the proposed model can newlinesave an average power of 28.15% and 64% in makespan. The results demonstrate newlinesignificant improvements in energy efficiency when compared to existing newlinealgorithms. This demonstrates the effectiveness and innovation of the proposed newlineapproach. The developed model dynamically newlinepredicts most suitable cores for individual workloads at runtime, thereby newlineoptimizing energy efficiency and overall system performance. The model newlineintegrates deep learning algorithms to predict core assignments for tasks along newlinewith scalability, cache management and performance asymmetry.