BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250626T233531Z
LOCATION:B302-B305
DTSTART;TZID=America/New_York:20241121T100000
DTEND;TZID=America/New_York:20241121T170000
UID:submissions.supercomputing.org_SC24_sess534_post136@linklings.com
SUMMARY:Poseidon: A Source-to-Source Translator for Holistic HPC Optimizat
 ion of Ocean Models on Regular Grids
DESCRIPTION:Maurice Brémond (French Institute for Research in Computer Sci
 ence and Automation (INRIA); Université Grenoble Alpes, France); Hugo Brun
 ie (Université Grenoble Alpes, France; French Institute for Research in Co
 mputer Science and Automation (INRIA)); Laurent Debreu (French Institute f
 or Research in Computer Science and Automation (INRIA); Université Grenobl
 e Alpes, France); Rupert W. Ford (Hartree Centre, Science and Technology F
 acilities Council (STFC), UK); Florian Lemarié (French Institute for Resea
 rch in Computer Science and Automation (INRIA); Université Grenoble Alpes,
  France); Anna Mittermair (Technical University of Munich); Andrew R. Port
 er (Hartree Centre, Science and Technology Facilities Council (STFC), UK);
  Julien Rémy (French Institute for Research in Computer Science and Automa
 tion (INRIA); Université Grenoble Alpes, France); Philippe Rosales and Mar
 tin Schreiber (Université Grenoble Alpes, France; French Institute for Res
 earch in Computer Science and Automation (INRIA)); Martin Schulz (Technica
 l University Munich); Sergi Siso (Hartree Centre, Science and Technology F
 acilities Council (STFC), UK); and Arthur Vidard (French Institute for Res
 earch in Computer Science and Automation (INRIA); Université Grenoble Alpe
 s, France)\n\nOcean simulation models often underperform on modern high-pe
 rformance computing (HPC) architectures, necessitating costly and time-con
 suming code rewrites.\n\nWe introduce Poseidon, an HPC-oriented source-to-
 source translator for Fortran-based fluid dynamics solvers used in ocean a
 nd weather models with regular grid structures. Poseidon aims to recover h
 igh-level information and semantics lost during the process of converting 
 numerics to source code.\n\nWe demonstrate Poseidon's approach using a res
 earch code implementing the 2D fast barotropic solver of full 3D ocean sim
 ulation models, which involves over 20 stencil-like kernels. Kernel fusion
 -based code optimization can already lead to a high combinatorial complexi
 ty.\n\nPreliminary results include various performance studies with and wi
 thout data flow graph-based modifications based on an exhaustive search fo
 r kernel fusion. Measurements show that Poseidon can generate optimized Fo
 rtran code.\n\nIn future work, Poseidon automatic code rewrite should help
  to: port existing code to GPU, hide process communications latency and ap
 ply automatic differentiation.\n\nRegistration Category: Tech Program Reg 
 Pass, Exhibits Reg Pass\n\nSession Chairs: Ayesha Afzal (Friedrich-Alexand
 er University, Erlangen-Nuremberg; Erlangen National High Performance Comp
 uting Center); Sally Ellingson (University of Kentucky); and Alan Sussman 
 (University of Maryland)\n\n
END:VEVENT
END:VCALENDAR