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:20230124T171527Z
LOCATION:C147-154
DTSTART;TZID=America/Chicago:20221114T103000
DTEND;TZID=America/Chicago:20221114T105200
UID:submissions.supercomputing.org_SC22_sess452_ws_pawatm109@linklings.com
SUMMARY:Extending OpenMP and OpenSHMEM for Efficient Heterogeneous Computi
 ng
DESCRIPTION:Workshop\n\nExtending OpenMP and OpenSHMEM for Efficient Heter
 ogeneous Computing\n\nLu, Tian, Curtis, Chapman\n\nHeterogeneous supercomp
 uting systems are becoming the mainstream thanks to their powerful acceler
 ators. However, the accelerators' special memory model and APIs increase t
 he development complexity, and calls for innovative programming model desi
 gns. To address this issue, OpenMP has added target offloading for portabl
 e accelerator programming, and MPI allows transparent send-receive of acce
 lerator memory buffers. Meanwhile, Partitioned Global Address Space (PGAS)
  languages like OpenSHMEM are falling behind for heterogeneous computing b
 ecause their special memory models pose additional challenges.\n\nWe propo
 se language and runtime interoperability extensions for both OpenMP and Op
 enSHMEM to enable portable remote access on GPU buffers, with minimal amou
 nt of code changes. Our modified runtime systems work in coordination to m
 anage accelerator memory, eliminating the need for staging communication b
 uffers. Comparing to the standard implementation, our extensions attain 6x
  point-to-point latency improvement, 1.3x better collective operation late
 ncy, 4.9x random access throughput, and up to 12.5% higher strong scalabil
 ity.\n\nSession Format: Recorded\n\nTag: Applications, Architectures, Hete
 rogeneous Systems, Hierarchical Parallelism, Parallel Programming Language
 s and Models, Performance, Performance Portability, Scientific Computing\n
 \nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR