Increasing the locality of a memory access profile is an interesting optimization problem, whose solution has implications on both performance and energy consumption of a processor-based system. High-locality memory access patterns are normally obtained through software-driven transformations, applied either at the source-code level or during compilation and machine-code generation. For some specific classes of systems, e.g., embedded systems, hardware-oriented, on-the-fly access pattern modification is also an option, as it trades generality in the applicability for better performance. This paper presents a class of algorithms for improving the locality of a given memory access profile through selective address relocation that require a very low-overhead custom hardware unit. Results obtained on a set of typical embedded applications show an average reduction in the complexity of the address relocation unit around 40%, yet with the best achievable increase in memory access locality.

Increasing the Locality of Memory Access Patterns by Low-Overhead Hardware Address Relocation / Macii, Alberto; Macii, E.; Poncino, Massimo. - 5:(2003), pp. 385-388. (Intervento presentato al convegno ISCAS-03: IEEE International Symposium on Circuits and Systems tenutosi a Bangkok, Thailandia nel Maggio 2003) [10.1109/ISCAS.2003.1206288].

Increasing the Locality of Memory Access Patterns by Low-Overhead Hardware Address Relocation

MACII, Alberto;E. MACII;PONCINO, MASSIMO
2003

Abstract

Increasing the locality of a memory access profile is an interesting optimization problem, whose solution has implications on both performance and energy consumption of a processor-based system. High-locality memory access patterns are normally obtained through software-driven transformations, applied either at the source-code level or during compilation and machine-code generation. For some specific classes of systems, e.g., embedded systems, hardware-oriented, on-the-fly access pattern modification is also an option, as it trades generality in the applicability for better performance. This paper presents a class of algorithms for improving the locality of a given memory access profile through selective address relocation that require a very low-overhead custom hardware unit. Results obtained on a set of typical embedded applications show an average reduction in the complexity of the address relocation unit around 40%, yet with the best achievable increase in memory access locality.
2003
0780377613
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

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/11583/1500535
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo