Precision tuning trades accuracy for speed and energy savings, usually by reducing the data width, or by switching from floating point to fixed point representations. However, comparing the precision across different representations is a difficult task. We present a metric that enables this comparison, and employ it to build a methodology based on Integer Linear Programming for tuning the data type selection. We apply the proposed metric and methodology to a range of processors, demonstrating an improvement in performance (up to 9 x) with a very limited precision loss (<2.8% for 90% of the benchmarks) on the PolyBench benchmark suite.

Architecture-aware Precision Tuning with Multiple Number Representation Systems

Cattaneo, Daniele;Chiari, Michele;Fossati, Nicola;Cherubin, Stefano;Agosta, Giovanni
2021-01-01

Abstract

Precision tuning trades accuracy for speed and energy savings, usually by reducing the data width, or by switching from floating point to fixed point representations. However, comparing the precision across different representations is a difficult task. We present a metric that enables this comparison, and employ it to build a methodology based on Integer Linear Programming for tuning the data type selection. We apply the proposed metric and methodology to a range of processors, demonstrating an improvement in performance (up to 9 x) with a very limited precision loss (<2.8% for 90% of the benchmarks) on the PolyBench benchmark suite.
2021
2021 58th ACM/IEEE Design Automation Conference (DAC)
978-1-6654-3274-0
File in questo prodotto:
File Dimensione Formato  
2021064384.pdf

accesso aperto

Descrizione: Post-print manuscript
: Post-Print (DRAFT o Author’s Accepted Manuscript-AAM)
Dimensione 362.97 kB
Formato Adobe PDF
362.97 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: https://hdl.handle.net/11311/1190472
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 13
  • ???jsp.display-item.citation.isi??? 3
social impact