@PHDTHESIS{ 2021:684446892, title = {Dynamic thermal management for noc-based many-core systems}, year = {2021}, url = "http://tede2.pucrs.br/tede2/handle/tede/9831", abstract = "Recent technology nodes enable to deploy billions of transistors in a small silicon area by replicating identical structures, resulting in many-core systems. However, power density may limit the amount of power the system can consume. A many-core at its maximum performance may lead to violate safe temperature definition and, consequently, result in reliability issues. Dynamic Thermal Management (DTM) techniques have been proposed to guarantee that many-core systems run at good performance without compromising reliability. DTM techniques rely on accurate temperature monitoring data. This Thesis reviews recent DTM works and proposes a new method to enable runtime temperature monitoring in a many-core system and new heuristics for thermal-aware application mapping, migration, and dynamic frequency and voltage scaling (DVFS) actuation, considering temperature and energy consumption. The state-of-art study on DTM techniques presents complex mechanisms for runtime actuation, focusing mainly on task mapping, and shows no concern about temperature monitoring methods, compromising the applicability in real systems. The stateof-art also presents works targeting dynamic reliability management (DRM), where the main objective is to ensure longer lifetime for the system, using the same actuation knobs used to control the temperature. This thesis also reviews some of the aging effects and analyses the lifetime reliability results for the proposed heuristics. The main contributions of this Thesis include: (i) a thermal estimation accelerator (TEA), (ii) a proportional, integral and derivative temperature management (PIDTM) heuristic; (iii) a temperature management heuristic having as cost function energy constraints (TMEC). TEA enabled accurate runtime temperature monitoring in the reference many-core, enabling the validation of the DTM proposals. PIDTM reduced up to 7.15% the overall peak temperature in a high workload scenario, while TMEC improved up to 82.9% in the system?s expected lifetime.", publisher = {Pontif?cia Universidade Cat?lica do Rio Grande do Sul}, scholl = {Programa de P?s-Gradua??o em Ci?ncia da Computa??o}, note = {Escola Polit?cnica} }