O uso do método do enxame de partículas na otimização da latência e consumo de energia de uma de rede-em-chip.

Abstract

With the development of applications for personal computers, servers and electronic devices, which require increasing computational power, came the need for technologies that improving both the physical performance such as logical. The network-on-chip, which is a set of routers and channels point-to-point interconnect to the core of an integrated system to support communication between these cores, appeared as an option for optimization of architectural hardware. In relation to software, there are several techniques that use intelligent algorithms, such as Particle Swarm, which was proposed by Kennedy and Eberhart in 1995, and an algorithm for optimization of stochastic simple concept, easy implementation, robustness to control parameters and computational efficiency during the optimization. Their heuristic is based on evolving a population (swarm), formed by individuals (particles), whose evolution is given by the speed. Each particle has a speed that allows the search through the space and position in space. This work aimed to optimize the latency and energy consumption of a network-on-chip, using two different topologies, the mesh and Toroid through a simulation using the method of particle swarm which, through the experiments, was efficient in this type of optimization.