ETD

Digital archive of theses discussed at the University of Pisa

 

Thesis etd-06132019-230957


Thesis type
Tesi di laurea magistrale
Author
UMANI, FEDERICO
URN
etd-06132019-230957
Thesis title
Modelling Simultaneous Multithreading Contribution to Performances and Power Consumption
Department
INFORMATICA
Course of study
INFORMATICA
Supervisors
relatore Dott. De Sensi, Daniele
Keywords
  • self-adaptive computing
  • power-aware computing
  • parallel computing
  • linear regression model
  • simultaneous multithreading
Graduation session start date
26/07/2019
Availability
Full
Summary
Energy demanding trend in Information and Communication Technologies makes power consumption an important metric in parallel computations. In this thesis we focus on the contribution brought by the Simultaneous Multithreading (SMT) technology both on performances and on power consumption. We build a linear regression model that integrates SMT and we show that it can predict both metrics more accurately with respect to another model that ignores SMT. This is tested both on offline data, by calculating the model error for a test set, and online by integrating it in Nornir, a framework for self-adaptive computing.
File