Modern embedded systems design increasingly couple hardware accelerators and processing units to obtain the right balance between energy efficiency and platform specialization. In this context High Level Synthesis (HLS) design flows can significantly reduce the effort needed to design these architectures. This thesis introduces a new HLS methodology that supports the synthesis of tightly coupled clusters of accelerators suitable to be used as building blocks for architecture design with the introduction of PorkSoC, a templated System on Chip (SoC) suitable for automatic generation. The proposed methodology introduces the concept of non-inlined function call and of function pointers to the traditional HLS methodology to synthesize cluster of accelerators able to call other accelerators to perform their computation without the mediation of a General Purpose Processor (GPP).

An architecture for function pointers and non inlined function calls in high level synthesis

MINUTOLI, MARCO
2012/2013

Abstract

Modern embedded systems design increasingly couple hardware accelerators and processing units to obtain the right balance between energy efficiency and platform specialization. In this context High Level Synthesis (HLS) design flows can significantly reduce the effort needed to design these architectures. This thesis introduces a new HLS methodology that supports the synthesis of tightly coupled clusters of accelerators suitable to be used as building blocks for architecture design with the introduction of PorkSoC, a templated System on Chip (SoC) suitable for automatic generation. The proposed methodology introduces the concept of non-inlined function call and of function pointers to the traditional HLS methodology to synthesize cluster of accelerators able to call other accelerators to perform their computation without the mediation of a General Purpose Processor (GPP).
ING - Scuola di Ingegneria Industriale e dell'Informazione
29-apr-2014
2012/2013
Tesi di laurea Magistrale
File allegati
File Dimensione Formato  
master-thesis.pdf

accessibile in internet per tutti

Descrizione: Thesis text
Dimensione 606.7 kB
Formato Adobe PDF
606.7 kB Adobe PDF Visualizza/Apri

I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10589/92065