BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230124T171526Z
LOCATION:C1-2-3
DTSTART;TZID=America/Chicago:20221115T083000
DTEND;TZID=America/Chicago:20221115T170000
UID:submissions.supercomputing.org_SC22_sess250_drs102@linklings.com
SUMMARY:Machine Learning for Memory Access Prediction and Data Prefetching
DESCRIPTION:Doctoral Showcase, Posters\n\nMachine Learning for Memory Acce
 ss Prediction and Data Prefetching\n\nZhang\n\nWith the rise of Big Data, 
 there has been a significant effort in increasing compute power through GP
 Us, TPUs, and heterogeneous architectures. As a result, the bottleneck of 
 applications is shifting toward memory performance. Prefetching techniques
  are widely used to hide memory latency and improve instructions per cycle
  (IPC). A data prefetching process is a form of speculation that looks at 
 memory access patterns to forecast the near future accesses and avoid cach
 e misses. Traditional hardware data prefetchers use pre-defined rules, whi
 ch are not powerful enough to adapt to the increasingly complex memory acc
 ess patterns from new workloads.\n\nWe hypothesize that a machine learning
 -based prefetcher can be developed to achieve high-quality memory access p
 rediction, leading to the improvement of IPC for a system. We develop seve
 ral optimizations for ML-based prefetching. First, we propose RAOP, a fram
 ework for RNN augmented offset prefetcher, in which RNN provides temporal 
 references for a spatial offset prefetcher, leading to the improvement of 
 IPC. Second, we propose C-MemMAP, which provides clusters for downstream m
 eta-models to balance the model size and prediction accuracy. We propose D
 M (delegated model) clustering method that learns latent patterns from lon
 g memory traces, which has significantly raised the prediction accuracy of
  the meta-models. Third, we propose TransFetch, an attention-based prefetc
 her that supports variable-degree prefetching by modeling prefetching as a
  multi-label classification problem. In addition, we propose ReSemble, a R
 einforcement Learning (RL) based adaptive ensemble framework that enables 
 multiple prefetchers to complement each other on hybrid applications and u
 pdates online.\n\nRegistration Category: Tech Program Reg Pass, Exhibits R
 eg Pass
END:VEVENT
END:VCALENDAR