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:20260522T150111Z
LOCATION:C2/3/4 Ballroom
DTSTART;TZID=America/Chicago:20181115T083000
DTEND;TZID=America/Chicago:20181115T170000
UID:submissions.supercomputing.org_SC18_sess324_post149@linklings.com
SUMMARY:GPU Acceleration at Scale with OpenPower Platforms in Code_Saturne
DESCRIPTION:Samuel Antao (IBM); Charles Moulinec (Science and Technology F
 acilities Council, UK); Yvan Fournier (EDF Research and Development); Robe
 rt Sawko, Malgorzata Zimon, and Christopher Thompson (IBM); Alex Skillen (
 Science and Technology Facilities Council, UK); Juan Uribe (EDF Research a
 nd Development); and David Emerson (Science and Technology Facilities Coun
 cil, UK)\n\nCode_Saturne is a widely used computational fluid dynamics sof
 tware package that uses finite-volume methods to simulate different kinds 
 of flows tailored to tackle multi-bilion-cell unstructured mesh simulation
 s. This class of codes has shown to be challenging to accelerate on GPUs a
 s they consist of many kernels and regular inter-process communication in 
 between. In this poster we show how template pack expansion with CUDA can 
 combine multiple kernels into a single one reducing launching latencies an
 d along with the specification of data environments help reduce host-devic
 e communication. We tested these techniques on ORNL Summit Supercomputer b
 ased on OpenPOWER platform delivering almost 3x speedup over CPU-only runs
  on 256 nodes. We also show how the latest generation NVLINK(TM) interconn
 ect available in POWER9(TM)improves scaling efficiency, enabling consisten
 t GPU acceleration with just 100K-cells per process.\n\nRegistration Categ
 ory: Tech Program Reg Pass, Exhibits Reg Pass\n\n
END:VEVENT
END:VCALENDAR
