Modular Modelling of an Embedded Mobile CPU-GPU Chip for Feature Estimation

Modélisation modulaire d'une puce CPU-GPU mobile intégrée pour des fins d'estimation de caractéristiques

Oussama Djedidi (1) , Mohand Arab Djeziri (1) , Nacer M'Sirdi (1) , Aziz Naamane (1)

1 SASV - Systèmes Automatisés à Structure Variable

This paper deals with the modelling of a CPU-GPUchip embedded in an Android phone. The model is used for the estimation of variables that characterise the operating state of System on Chip (SoC). The proposed model is built to demonstrate the causal relationships between the variables, through its interconnected structure of subsystems. This structure allows the extension of other components or the easy exchange of subsystems in the case of a change in components or operating mode. The model developed here requires no additional instrumentation—other than the one present on the phone—which facilitates its implementation. It is used for the estimation of the state of the system and can also be used for monitoring and behaviour prediction. The model is validated and the results are promising for further implementation.

Mots clés: Embedded Chips Central Processing Units; Graphics Processing Units; Modelling Simulation

Domaines: Electronique; Machine Learning [stat.ML]; Automatique / Robotique

Oussama Djedidi, Mohand Arab Djeziri, Nacer M'Sirdi, Aziz Naamane. Modular Modelling of an Embedded Mobile CPU-GPU Chip for Feature Estimation. 14th International Conference on Informatics in Control, Automation and Robotics, Jul 2017, Madrid, Spain. pp.338-345, ⟨10.5220/0006470803380345⟩. ⟨hal-01770233⟩