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:20211207T055343Z
LOCATION:Online
DTSTART;TZID=America/Chicago:20211114T143000
DTEND;TZID=America/Chicago:20211114T150000
UID:submissions.supercomputing.org_SC21_sess432_ws_waccpd103@linklings.com
SUMMARY:Extending OpenMP for Machine Learning-Driven Adaptation
DESCRIPTION:Workshop\n\nExtending OpenMP for Machine Learning-Driven Adapt
 ation\n\nLiao, Wang, Georgakoudis, de Supinski, Yan...\n\nOpenMP 5.0 intro
 duced themetadirectivedirective to sup-port compile-time selection from a 
 set of directive variants based onOpenMP context. OpenMP 5.1 extended cont
 ext information to include user-defined conditions that enable user-guided
  runtime adaptation. How-ever, defining conditions that capture the comple
 x interactions between applications and hardware platforms to select an op
 timized variant is challenging for programmers. This paper explores a nove
 l approach to automate runtime adaptation through machine learning. In par
 ticular,we design a new omp declare adaptation directive and its associate
 d clauses to describe semantics for model-driven adaptation and also devel
 op a prototype source-to-source transformation tool for evaluating our run
 time adaptation approach. Leveraging an existing runtime library for tunin
 g, we design a small set of API functions to support our source-to-source 
 compiler transformations. Our evaluation, using the Smith-Waterman algorit
 hm as a use-case, demonstrates that the proposed adap-tive OpenMP extensio
 n automatically chooses the code variants that deliver the best performanc
 e in heterogeneous platforms that consists ofCPU and GPU processing capabi
 lities. Using decision tree models for tuning has an accuracy of up to 93.
 1% in selecting the optimal variant,with negligible runtime overhead.\n\nT
 ag: Online Only, Accelerator-based Architectures, Parallel Programming Lan
 guages and Models, Performance, State of the Practice\n\nRegistration Cate
 gory: Workshop Reg Pass
END:VEVENT
END:VCALENDAR