BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20200129T163558Z
LOCATION:704-706
DTSTART;TZID=America/Denver:20191117T153000
DTEND;TZID=America/Denver:20191117T160000
UID:submissions.supercomputing.org_SC19_sess114_ws_prot106@linklings.com
SUMMARY:Automatic Instrumentation Refinement for Empirical Performance Mod
 eling
DESCRIPTION:Workshop\n\nAutomatic Instrumentation Refinement for Empirical
  Performance Modeling\n\nLehr, Calotoiu, Bischof, Wolf\n\nThe analysis of 
 runtime performance is important during the development and throughout the
  life cycle of HPC applications.  One important objective in performance a
 nalysis is to identify regions in the code that show significant runtime i
 ncrease with larger problem sizes or more processes.  One approach to iden
 tify such regions is to use empirical performance modeling, i.e., building
  performance models based on measurements.  While the modeling itself has 
 already been streamlined and automated, the generation of the required mea
 surements is time consuming and tedious.  In this paper, we propose an app
 roach to automatically adjust the instrumentation to reduce overhead and f
 ocus the measurements to relevant regions, i.e., such that show increasing
  runtime with larger input parameters or increasing number of MPI ranks.  
 Our approach employs Extra-P to generate performance models, which it then
  uses to extrapolate runtime and, finally, decide which functions should b
 e kept for measurement.  Also, the analysis expands the instrumentation, b
 y heuristically adding functions based on static source-code features.  We
  evaluate our approach using benchmarks from SPEC CPU 2006, SU2, and paral
 lel MILC.  The evaluation shows that our approach can filter functions of 
 little interest and generate profiles that contain mostly relevant regions
 .  For example, the overhead for SU2 can be improved automatically from 20
 0% to 11% compared to filtered Score-P measurements.\n\nTag: Workshop Reg 
 Pass, Performance, Programming Systems, Visualization\n\nRegistration Cate
 gory: Workshop Reg Pass, Performance, Programming Systems, Visualization
URL:https://sc19.supercomputing.org/presentation/?id=ws_prot106&sess=sess1
 14
END:VEVENT
END:VCALENDAR