Title:
Data Prefetching Using Off-Line Learning
Data Prefetching Using Off-Line Learning
Author(s)
Kim, Jinwoo
Palem, Krishna V.
Wong, Weng Fai
Palem, Krishna V.
Wong, Weng Fai
Advisor(s)
Editor(s)
Collections
Supplementary to
Permanent Link
Abstract
An important technique for alleviating the memory bottleneck is data prefetching. Data
prefetching solutions ranging from insertion of prefetch instructions by means of program
analysis to strictly hardware prefetch mechanisms have been proposed. The former,
however, is less successful for pointer intensive applications. In this paper, we propose a
hardware solution that utilizes off-line learning algorithms. In essence, a sample trace of
the application is fed into various off-line learning schemes. The results from these
schemes are then loaded into a prefetching hardware at the appropriate point in the
execution of the application to drive the prefetching. We propose a general architecture
and scheme for such a process and report on the results of some of the experiments we
performed.
Sponsor
Date Issued
2001
Extent
267237 bytes
Resource Type
Text
Resource Subtype
Technical Report