Approximate Computing is revealing a new design paradigm which trades algorithms precision off for enhancing performance parameters, commonly energy consumption and computation time. Applications which are characterized by the inherent resiliency property tolerate some quality loss, w. r. t. the optimal result. The approximation is accomplished by combining substitutions of fully-precise block operations with inaccurate ones. However, exploring every possible approximate variant of an algorithm would be extremely costly due to countless configurations. IIDEAA, a design exploration tool for approximate computing algorithms, introduced a branch and bound exploration approach to make it affordable. In this paper, we enhance the IIDEAA B & B exploration approach by introducing a pruning technique, which significantly reduces the design solution space to explore. We demonstrate the effectiveness of approach by comparing the execution of approximating campaigns over some algorithms employing proposed pruning rules.

A pruning technique for B & B based design exploration of approximate computing variants / Barbareschi, M.; Iannucci, F.; Mazzeo, A.. - 2016-:(2016), pp. 707-712. (Intervento presentato al convegno 15th IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2016 tenutosi a usa nel 2016) [10.1109/ISVLSI.2016.110].

A pruning technique for B & B based design exploration of approximate computing variants

Barbareschi M.;
2016

Abstract

Approximate Computing is revealing a new design paradigm which trades algorithms precision off for enhancing performance parameters, commonly energy consumption and computation time. Applications which are characterized by the inherent resiliency property tolerate some quality loss, w. r. t. the optimal result. The approximation is accomplished by combining substitutions of fully-precise block operations with inaccurate ones. However, exploring every possible approximate variant of an algorithm would be extremely costly due to countless configurations. IIDEAA, a design exploration tool for approximate computing algorithms, introduced a branch and bound exploration approach to make it affordable. In this paper, we enhance the IIDEAA B & B exploration approach by introducing a pruning technique, which significantly reduces the design solution space to explore. We demonstrate the effectiveness of approach by comparing the execution of approximating campaigns over some algorithms employing proposed pruning rules.
2016
978-1-4673-9039-2
A pruning technique for B & B based design exploration of approximate computing variants / Barbareschi, M.; Iannucci, F.; Mazzeo, A.. - 2016-:(2016), pp. 707-712. (Intervento presentato al convegno 15th IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2016 tenutosi a usa nel 2016) [10.1109/ISVLSI.2016.110].
File in questo prodotto:
Non ci sono file associati a questo prodotto.

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/11588/915806
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 17
  • ???jsp.display-item.citation.isi??? 14
social impact