logo SBA

ETD

Archivio digitale delle tesi discusse presso l’Università di Pisa

Tesi etd-05312021-190228


Tipo di tesi
Tesi di laurea magistrale
Autore
RE, WALTER ENZO
URN
etd-05312021-190228
Titolo
Implementation of an Energy Aware Scheduler for arm big.LITTLE architecture
Dipartimento
INGEGNERIA DELL'INFORMAZIONE
Corso di studi
EMBEDDED COMPUTING SYSTEMS
Relatori
relatore Prof. Abeni, Luca
correlatore Prof. Cucinotta, Tommaso
Parole chiave
  • arm big.LITTLE
  • kernel
  • scheduler
Data inizio appello
21/06/2021
Consultabilità
Completa
Riassunto
The growth of performance in modern CPUs has brought an increase in power consumption, forcing CPU manufacturers to move from a single powerful core to platforms with multiple cores. Nowadays multicore processors are the standard and this required to design new multi-core CPU schedulers. There are two solutions: partitioning scheduling, which tries to reduce a multi-core system to multiple uni-core subsystems and global scheduling which selects, each time, a new queue for each task. While most of the desktop PCs are based on Symmetric Multicores, mobile systems are generally based on Asymmetric Multiprocessors such as the ARM big.LITTLE architecture. These architectures contain powerful and power angry cores together with slower but power-saving ones. Since global scheduling has been developed for a homogeneous platform, it could not perform well on a solution of that type. Hence, ARM has implemented the Energy Aware Scheduler in Android and in the mainline Linux Kernel. This scheduler uses global scheduling to find the best CPU to minimize the overall energy consumption in the system. The purpose of this thesis is to implement Energy Aware Scheduling also for the deadline scheduling class. The first step was building up an adaptively partitioned version of SCHED_DEADLINE. The second step was bringing arm EAS implementation and adapt it for this new scheduler. Finally, another Energy Aware Scheduling algorithm was proposed.
File