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:20230124T171521Z
LOCATION:C144-145
DTSTART;TZID=America/Chicago:20221114T143500
DTEND;TZID=America/Chicago:20221114T145000
UID:submissions.supercomputing.org_SC22_sess462_ws_ai4s107@linklings.com
SUMMARY:Pattern-Based Autotuning of OpenMP Loops Using Graph Neural Networ
 ks
DESCRIPTION:Workshop\n\nPattern-Based Autotuning of OpenMP Loops Using Gra
 ph Neural Networks\n\nDutta, Alcaraz, Tehrani Jamsaz, Sikora, Cesar...\n\n
 Frequently occurring code and design patterns in scientific applications a
 re often used for parallelizing serial code.  But, identifying these patte
 rns is difficult.  We propose using Graph Neural Networks for modeling cod
 e flow graphs to identify patterns in such parallel code.  Additionally, i
 dentifying the best runtime parameters for parallel code is also challengi
 ng.  We propose a pattern-guided deep learning based tuning approach, to i
 dentify the best runtime parameters for OpenMP loops.  We validate our hyp
 othesis on 20 different applications from Polybench, and STREAM benchmark 
 suites.  Our approach identifies patterns with an accuracy of 91%.  We val
 idate the usefulness of using patterns for auto-tuning, on tuning the numb
 er of threads, scheduling policies and chunk size on a single-socket syste
 m, and the thread count and affinity on a multi-socket machine.  We achiev
 e geometric mean speedups of 1.1X and 4.7X respectively over default OpenM
 P configurations, compared to brute-force speedups of 1.27X and 4.93X resp
 ectively.\n\nSession Format: Recorded\n\nRegistration Category: Workshop R
 eg Pass
END:VEVENT
END:VCALENDAR