Controle adaptativo para atendimento a requisitos de aplicações em MPSoCS

• Ano de defesa: 2013
• Autor(a) principal: Madalozzo, Guilherme Afonso
• Orientador(a): Moraes, Fernando Gehm
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação
• Departamento: Faculdade de Informáca
• País: BR
• Palavras-chave em Português: INFORMÁTICA; MULTIPROCESSADORES; ALGORITMOS
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://tede2.pucrs.br/tede2/handle/tede/5218

The growing number of manufactured transistors in embedded systems follows the trend of Moore s Law, which states that every eighteen months the number of transistors on integrated circuits doubles, while its cost remains constant. Another important issue in embedded systems is that applications with more than one processor are increasingly present in market. These devices with several processing elements are named MPSoCs (Multiprocessor System-on-Chip). MPSoCs enables the development of complex systems, together with high performance. Applications executing in MPSoC have constraints to be respected. To meet these constraints, management techniques and resources adaptability should be researched and developed. This work presents the development and evaluation of adaptive management techniques that enable applications executing in MPSoCs to meet their performance requirements. The MPSoC management uses monitoring techniques, which evaluate applications constraints, as throughput and latency. When violations are detected by the monitoring infrastructure, adaptive techniques are executed. In the scope of this work, two techniques were developed: dynamic change in the priority scheduling of tasks and task migration. The evaluation of the proposed techniques is carried out using the HeMPS MPSoC, with centralized and distributed resource management. Results show that, regardless the resource management technique adopted, the proposed adaptive techniques decrease latency and jitter, without affecting the total execution time of applications. With performed adaptive techniques the total execution time wasn t penalized, in presented experiments increased 7%.