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:20260522T150118Z
LOCATION:D166
DTSTART;TZID=America/Chicago:20181112T120000
DTEND;TZID=America/Chicago:20181112T123000
UID:submissions.supercomputing.org_SC18_sess173_ws_espm106@linklings.com
SUMMARY:Automatic Generation of High-Order Finite-Difference Code with Tem
 poral Blocking for Extreme-Scale Many-Core Systems
DESCRIPTION:Hideyuki Tanaka (ExaScaler Inc), Youhei Ishihara (Kyoto Univer
 sity), Ryo Sakamoto and Takashi Nakamura (PEZY Computing), Yasuyuki Kimura
  (ExaScaler Inc), Keigo Nitadori and Miyuki Tsubouchi (Riken Center for Co
 mputational Science), and Jun Makino (Kobe University)\n\nIn this paper we
  describe the basic idea, implementation and achieved performance of our D
 SL for stencil computation, Formura, on systems based on PEZY-SC2 many-cor
 e processor. Formura generates, from high-level description of the differe
 ntial equation and simple description of finite-difference stencil, the en
 tire simulation code with MPI parallelization with overlapped communicatio
 n and calculation, advanced temporal blocking and parallelization for many
 -core processors.  Achieved performance is 4.78 PF, or 21.5% of the theore
 tical peak performance for an explicit scheme for compressive CFD, with th
 e accuracy of fourth-order in space and third-order in time. For a slightl
 y modified implementation of the same scheme, efficiency was slightly lowe
 r (17.5%)  but actual calculation time per one timestep was faster by 25%.
   Temporal blocking improved the performance by up to 70%.  Even though th
 e B/F number of PEZY-SC2 is low, around 0.02, we have achieved the efficie
 ncy comparable to those of highly optimized CFD codes on machines with muc
 h higher memory bandwidth such as K computer. We have demonstrated that au
 tomatic generation of the code with temporal blocking is a quite effective
  way to make use of very large-scale machines with low memory bandwidth fo
 r large-scale CFD calculations.\n\nTag: Accelerators, Exascale, Parallel P
 rogramming Languages, Libraries, and Models\n\nRegistration Category: Work
 shop Reg Pass\n\nSession Chairs: Dhabaleswar K. (DK) Panda (The Ohio State
  University), Karl Schulz (Advanced Micro Devices (AMD) Inc), and Hari Sub
 ramoni (The Ohio State University)\n\n
END:VEVENT
END:VCALENDAR