BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Stream Chaining: Exploiting Multiple Levels of Correlation in Data
  Prefetching - Marcelo Cintra
DTSTART:20110526T100000Z
DTEND:20110526T110000Z
UID:TALK31462@talks.cam.ac.uk
CONTACT:Robert Mullins
DESCRIPTION:Data prefetching has long been an important technique to amort
 ize the effects of the memory wall\, and is likely to remain so in the cur
 rent era of multi-core systems. Most prefetchers operate by identifying pa
 tterns and correlations in the miss address stream. Separating streams acc
 ording to the memory access instruction that generates the misses is an ef
 fective way of filtering out spurious addresses from predictable streams. 
 On the other hand\, by localizing streams based on the memory access instr
 uctions\, such prefetchers both lose the complete time sequence informatio
 n of misses and can only issue prefetches for a single memory access instr
 uction at a time.\n\nIn this talk we introduce a novel class of prefetcher
 s based on the idea of linking various localized streams into predictable 
 chains of missing memory access instructions such that the prefetcher can 
 issue prefetches along multiple streams. In this way the prefetcher is not
  limited to prefetching deeply for a single missing memory access instruct
 ion but can instead\nadaptively prefetch for other memory access instructi
 ons closer in time. Experimental results show that the proposed prefetcher
  consistently achieves better performance than a state-of-the-art prefetch
 er - 10% on average\, being only outperformed in very few cases and then b
 y only 2%\, and outperforming that prefetcher by as much as 55% - while co
 nsuming the same amount of memory bandwidth.
LOCATION:Lecture Theatre 1\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR