Considering the energy-cap problem in batterypowered devices, DVFS and power gating represent the defacto state-of-the-art actuators. However, the limited margin to reduce the operating voltage, the impossibility to massively integrate such actuators on-chip and their actuation latency force a revision of such design methodologies. We present an all-digital architecture and a design methodology to effectively manage the energy-cap problem for CPUs and accelerators. Two quality metrics are put forward to capture the performance loss and the energy budget violations. We employed a vector processor supporting 4 hardware threads as representative usecase. Results show an average performance loss and energy cap violations limited to 2.9% and 3.8%, respectively. Compared to solutions employing the DFS actuator, our all-digital architecture improves the energy-cap violations by 3x with a similar performance loss.

All-digital energy-constrained controller for general-purpose accelerators and CPUs

Davide Zoni;Luca Cremona;William Fornaciari
2019

Abstract

Considering the energy-cap problem in batterypowered devices, DVFS and power gating represent the defacto state-of-the-art actuators. However, the limited margin to reduce the operating voltage, the impossibility to massively integrate such actuators on-chip and their actuation latency force a revision of such design methodologies. We present an all-digital architecture and a design methodology to effectively manage the energy-cap problem for CPUs and accelerators. Two quality metrics are put forward to capture the performance loss and the energy budget violations. We employed a vector processor supporting 4 hardware threads as representative usecase. Results show an average performance loss and energy cap violations limited to 2.9% and 3.8%, respectively. Compared to solutions employing the DFS actuator, our all-digital architecture improves the energy-cap violations by 3x with a similar performance loss.
Energy-constrained design, low power, digital design, RTL design, multi-core, power management
File in questo prodotto:
File Dimensione Formato  
esl2019.pdf

accesso aperto

Descrizione: camera ready
: Pre-Print (o Pre-Refereeing)
Dimensione 270.57 kB
Formato Adobe PDF
270.57 kB Adobe PDF Visualizza/Apri
ES_Controller_08703179.pdf

Accesso riservato

Descrizione: versione pubblicata
: Publisher’s version
Dimensione 786.28 kB
Formato Adobe PDF
786.28 kB Adobe PDF   Visualizza/Apri

I documenti in IRIS 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: http://hdl.handle.net/11311/1085134
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 10
  • ???jsp.display-item.citation.isi??? 8
social impact