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:C155
DTSTART;TZID=America/Chicago:20221113T143800
DTEND;TZID=America/Chicago:20221113T150000
UID:submissions.supercomputing.org_SC22_sess428_ws_p3hpc127@linklings.com
SUMMARY:Piper: Pipelining OpenMP Offloading Execution through Compiler Opt
 imization for Performance
DESCRIPTION:Workshop\n\nPiper: Pipelining OpenMP Offloading Execution thro
 ugh Compiler Optimization for Performance\n\nParasyris, Georgakoudis, Doer
 fert, Laguna, Scogland\n\nOpenMP offload improves the application developm
 ent complexity of HPC GPU codes and provides portability. A source of poor
  performance is the lockstep execution of data transfers and computation. 
 Overlapping these operations can provide significant performance gains. Ho
 wever, the developer must manually slice data transfers and kernel executi
 on, and efficiently schedule these operations for execution, which is a ha
 rd and error-prone task.\n\nWe propose Piper, an automatic mechanism for O
 penMP offload to perform overlapping. Piper statically analyzes offload ke
 rnels and associates computations with memory locations. The extended runt
 ime system exploits this analysis information, divides a kernel into indep
 endent sub-tasks, and schedules them for pipelined execution for overlappi
 ng. At any point in time, Piper also controls the coarseness and number of
  sub-tasks executed. By doing so, Piper allows the execution of kernels wh
 ose memory requirements exceed the GPU device memory. Piper speeds up exec
 ution up to 2.67× compared to  OpenMP-offload execution.\n\nSession Format
 : Recorded\n\nTag: Performance Portability\n\nRegistration Category: Works
 hop Reg Pass
END:VEVENT
END:VCALENDAR
