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:20230124T171520Z
LOCATION:C143-149
DTSTART;TZID=America/Chicago:20221113T160000
DTEND;TZID=America/Chicago:20221113T163000
UID:submissions.supercomputing.org_SC22_sess433_ws_prot102@linklings.com
SUMMARY:Performance Debugging and Tuning of Flash-X with Data Analysis Too
 ls
DESCRIPTION:Workshop\n\nPerformance Debugging and Tuning of Flash-X with D
 ata Analysis Tools\n\nHuck, Wu, Dubey, Georgiadou, Harris...\n\nState-of-t
 he-art multiphysics simulations running on large scale leadership computin
 g platforms have many variables contributing to their performance and scal
 ing behavior. We recently encountered an interesting performance anomaly i
 n Flash-X, a multiphysics multicomponent simulation software, when charact
 erizing its performance behavior on several large-scale platforms. The ano
 maly was tracked down to the interaction between the use of dynamic alloca
 tion of scratch data and data locality in the cache hierarchy. In this pap
 er we present the details of unexpected performance variability we encount
 ered, the extensive analysis using the performance measurement tool TAU to
  collect the data and Python data analysis libraries to explore the data, 
 and our insights from this experience. In the process, we discovered and r
 emoved or mitigated two additional performance limiting bottlenecks for pe
 rformance tuning.\n\nSession Format: Recorded\n\nRegistration Category: Wo
 rkshop Reg Pass
END:VEVENT
END:VCALENDAR