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:20260522T150119Z
LOCATION:D161
DTSTART;TZID=America/Chicago:20181116T103000
DTEND;TZID=America/Chicago:20181116T105000
UID:submissions.supercomputing.org_SC18_sess153_ws_espt103@linklings.com
SUMMARY:PARLOT: Efficient Whole-Program Call Tracing for HPC Applications
DESCRIPTION:Saeed Taheri (University of Utah), Sindhu Devale (Texas State 
 University), Ganesh Gopalakrishnan (University of Utah), and Martin Burtsc
 her (Texas State University)\n\nThe complexity of HPC software and hardwar
 e is quickly increasing. As a consequence, the need for efficient executio
 n tracing to gain insight into HPC application behavior is steadily growin
 g. Unfortunately, available tools either do not produce traces with enough
  detail or incur large overheads. An efficient tracing method that overcom
 es the tradeoff between maximum information and minimum overhead is theref
 ore urgently needed. This paper presents such a method and tool, called Pa
 rLoT, with the following key features. (1) It describes a technique that m
 akes low-overhead on-the-fly compression of whole-program call traces feas
 ible. (2) It presents a new, highly efficient, incremental trace-compressi
 on approach that reduces the trace volume dynamically, which lowers not on
 ly the needed bandwidth but also the tracing overhead. (3) It collects all
  caller/callee relations, call frequencies, call stacks, as well as the fu
 ll trace of all calls and returns executed by each thread, including in li
 brary code. (4) It works on top of existing dynamic binary instrumentation
  tools, thus requiring neither source-code modifications nor recompilation
 . (5) It supports program analysis and debugging at the thread, thread-gro
 up, and program level. \n\nThis paper establishes that comparable capabili
 ties are currently unavailable. Our experiments with the NAS parallel benc
 hmarks running on the Comet supercomputer with up to 1,024 cores show that
  ParLoT can collect whole-program function-call traces at an average traci
 ng bandwidth of just 56 kB/s per core.\n\nTag: Performance, Productivity\n
 \nRegistration Category: Workshop Reg Pass\n\nSession Chairs: Martin Schul
 z (Technical University of Munich, Leibniz Supercomputing Centre (LRZ)) an
 d Felix Wolf (Technical University of Darmstadt)\n\n
END:VEVENT
END:VCALENDAR