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:20211207T055408Z
LOCATION:Online
DTSTART;TZID=America/Chicago:20211114T153100
DTEND;TZID=America/Chicago:20211114T160000
UID:submissions.supercomputing.org_SC21_sess426_ws_mchpc102@linklings.com
SUMMARY:Performance and Energy Improvement of the ECP Proxy App SW4lite un
 der Various Workloads
DESCRIPTION:Workshop\n\nPerformance and Energy Improvement of the ECP Prox
 y App SW4lite under Various Workloads\n\nWu, Taylor, Lan\n\nEnergy efficie
 nt execution of scientific applications requires insight into how HPC syst
 ems affect the performance and energy of the applications. In this paper, 
 we conduct experiments to evaluate the performance of SW4lite under variou
 s workloads with two different memory modes on Cray XC40 Theta at Argonne 
 National Laboratory. We use MuMMI and ensemble learning to build the perfo
 rmance-counter-based performance and power models, and we identify perform
 ance and power bottlenecks based on the insights from these performance an
 d power models. Then we improve the performance and energy of SW4lite with
  a focus on memory-centric optimizations and code modifications. The exper
 imental results show that our performance counter-guided application optim
 ization strategies result in up to 26.97% performance improvement and up t
 o 19.44% energy saving for SW4lite with various workloads on up to 16,384 
 cores.\n\nTag: Online Only, Architectures, Memory Systems, Parallel Progra
 mming Languages and Models, System Software and Runtime Systems\n\nRegistr
 ation Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR