Gerenciamento de energia em processadores com o uso de lógica fuzzy.

Abstract

General purpose computers are used in various environments, such as homes, entertainment centers, offices or in applications where high processing power is not required. In these applications, power management in their processors allows them to effectively reduce power consumption without altering the processor's performance in performing the requested tasks. The main objective of this work was the development of a software that allows processing the measurements of workload and controlling the use of the processor through the variation of the operating frequency of the central processing unit. For the use of Fuzzy logic, models for the use of processors were developed to be used in the implementation of other types of controllers and an experimental platform that allows changing the frequencies in the cores of the processors and performing the acquisition of data, such as the level of use of the processor, the power and energy consumed. For medium and low intensity workloads, the manager proposed in this work was able to promote reductions in the order of 20% in energy consumption. The use of Fuzzy logic requires little energy consumption and thus has little impact (0.93% of the energy consumed) on energy management. It was concluded that the use of learning and pattern recognition techniques such as Fuzzy logic proved to be effective in the development of models for the use of processors, as well as in the design and implementation of the control law.