New URL for NEMO forge!   http://forge.nemo-ocean.eu

Since March 2022 along with NEMO 4.2 release, the code development moved to a self-hosted GitLab.
This present forge is now archived and remained online for history.
WorkingGroups/HPC – NEMO
wiki:WorkingGroups/HPC

Version 4 (modified by smasson, 9 years ago) (diff)

--

TOC(heading=NEMO_HPC,NEMO_HPC/*, depth=1)?

NEMO HPC

Working group leader (and responsible for wiki pages) : Sébastien Masson.


Members of the Working group:

  • Sébastien Masson
  • Italo Epicoco
  • Silvia Mocavero
  • Marie-Alice Foujols
  • Jason Holt
  • Gurvan Madec
  • Mondher Chekki

Objectives:

  • make short term recommendations for improving the performance of the existing system
  • propose criteria for a taking decisions at Gateway 2025 regarding HPC.
  • provide more detail on Gung-Ho (esp. regarding its implications for mesh discretization)
  • identify other possible strategies and approaches for evolutions in the long term.
  • define a simple configuration (with IO and complex geometry) that will serve as a proof of concept for validating the proposed approach for the future system.

Some ideas...:

A strong improvement of NEMO scalability is needed to be able to take advantage of the new machines. This means a deep review/rewrite of NEMO code at some point in the futur (beyond 5 years). At the same time, we already know that CMIP7 won't use an ocean model that has not been strongly tested and validated and will stick to a NEMO model close to the existing one.
This means that we need to:

1) keep improving the current structure of NEMO so it works quite efficiently for almost 10 more years (until the end of CMPI7).
2) start to work on a new structure that would be tested from year 5 and fully tested and validated for CMIP8 in about 10 years.

Based on this, we propose to divide the work according to 3 temporal window

0-3 years: improvements with existing code: Do basic optimisation work:

1) reduce the number of communications: remove useless communications (a lot) 2) reduce the number of communications: do less and bigger communications (group communications, use large halo)
3) introduce asynchronous communications 3) check code vectorization (SIMD instructions)

0-5 years: improvements through the introduction of OpenMP:

work initialed by CMCC. "Basic" implementation such as tiling may be efficient with many cores processors? test different way to find new sources of parallelism.

OpenMP4, OpenACC?

beyond 5 years:

GungHo? or not GungHo?, that is the question...

Agenda:

For the next 2 years, as a start, a workshop to be organized in 2015 on “NEMO in 2025 : routes toward multi-resolution approaches”.

Attachments (5)

Download all attachments as: .zip