MetOffice / MetOffice/lfric_core

Halo exchange hangs for high-order W0 field on multigrid mesh in high-res runs

Open
#239 0 comments 0 reactions 0 assignees View on GitHub
bug
Dominant language
Fortran
Stars
26
Forks
73
Avg merge
1d 8h
Merged PRs (30d)
4

Description

Orography is set in gungho by specifying the `surface_altitude` field, which is a 2D W0 field. This field is then used to modify the vertical values of the `chi` fields.

The `chi` fields are never halo exchanged, and their modification is performed up to the full mesh halo depth. But to do this, we must perform a halo exchange on the `surface_altitude` field up to its full halo depth.

For large numbers of procs (e.g. 2000), when I increase the order of finite element for the `surface_altitude` field, I have found that the halo exchange hangs for the coarsest multigrid mesh and does not complete.

I have a workaround in #238 which keeps the coordinate order for multigrid meshes at 1

Contributor guide

No contributing guide indexed for this repository

Research direction

Start with the halo exchange for the surface_altitude field on the coarsest multigrid mesh, focusing on high-order finite elements and high-resolution runs. Compare the behavior with the workaround in issue #238; done means the exchange completes reliably at large processor counts.

Written by the indexing model from the issue text.

Assessment

Tech stack
fortran
Domain
distributed-systems
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.