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:D222
DTSTART;TZID=America/Chicago:20221114T144000
DTEND;TZID=America/Chicago:20221114T150000
UID:submissions.supercomputing.org_SC22_sess443_ws_xloop106@linklings.com
SUMMARY:Toward Increasing Trust in Exascale Simulations
DESCRIPTION:Workshop\n\nToward Increasing Trust in Exascale Simulations\n\
 nBen Khalifa, Li, Laguna, Martel, Gopalakrishnan\n\nIn recent decades, Hig
 h Performance Computing (HPC) and simulations have become determinant in m
 any areas of engineering and science. Since many HPC applications rely ext
 ensively on floating-point arithmetic operations,  many kinds of numerical
  errors can be introduced during  the program execution, leading to instab
 ility or reproducibility problems. One kind of these error sources is  can
 cellation which produces inaccurate results  when two  nearby numbers are 
 subtracted. In this article, we present Candy, a new  dynamic library  tha
 t detects cancellations in numerical codes. Our method  computes the numbe
 r of significant bits of floating-point numbers by attaching a shadow valu
 e in higher precision to each number. This helps  to detect in an accurate
  way if a program suffers from cancellation problems and thus to increase 
 the trust in large-scale HPC applications and exascale simulations. We eva
 luate Candy over a set of real-world numerical applications. Also, we comp
 are Candy against the  state-of-art tool FPChecker.\n\nSession Format: Rec
 orded\n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR