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:20260422T000712Z
LOCATION:601
DTSTART;TZID=America/Denver:20231112T110000
DTEND;TZID=America/Denver:20231112T113000
UID:submissions.supercomputing.org_SC23_sess426_ws_llvmf108@linklings.com
SUMMARY:Precision and Performance Analysis of C Standard Math Library Func
 tions on GPUs
DESCRIPTION:Anton Rydahl (Lawrence Livermore National Laboratory (LLNL)), 
 Joseph Huber (Advanced Micro Devices (AMD) Inc), and Ethan Luis McDonough 
 and Johannes Doerfert (Lawrence Livermore National Laboratory (LLNL))\n\nW
 ith the advent of GPU computing, executing large program sections on accel
 erators has become increasingly important.  Efforts are being made to supp
 ort the C standard library, LIBC, on GPUs via LLVM machinery.  Therefore, 
 the C standard math library, LIBM, must be supported on GPUs.  So far, LLV
 M frontends, such as Clang,  have relied on GPU vendor implementations of 
 LIBM functionality wrapped into (mostly) LIBM-compatible forwarding functi
 ons.\n\nWe propose a novel LIBM for GPUs reusing a collection of LLVM targ
 et-agnostic implementations and built-ins alongside vendor implementations
  of most single and double-precision floating point math functions.  Our a
 pproach allows selecting between individual implementations based on the G
 PU target as opposed to the current approach, which serves only the single
  third-party library implementation.  Our extensive numerical analysis hig
 hlights the various implementations' differences in performance and precis
 ion.  Our solution allows users to choose the implementation that maximize
 s speed while meeting their specific precision requirements.\n\nTag: Compi
 lers, Heterogeneous Computing, Performance Optimization\n\nRegistration Ca
 tegory: Workshop Reg Pass\n\nSession Chairs: James Brodman (AMD); Ryan Kab
 rick (Tactical Computing Laboratories LLC, University of Delaware); Patric
 k S. McCormick (Los Alamos National Laboratory (LANL)); and Alexis Perry-H
 olby (Los Alamos National Laboratory (LANL))\n\n
END:VEVENT
END:VCALENDAR