MetOffice / MetOffice/lfric_apps
investigate mpi-parts mpi-parts_xios ratio
- Dominant language
- Fortran
- Stars
- 31
- Forks
- 118
- Avg merge
- 2d 4h
- Merged PRs (30d)
- 8
Description
#470 && #519 introduced a seeming constraint for large weekly jobs that
* the number of `mpi_parts` now needs to be exactly divisible by the number of `mpi_parts_xios`
* it is not at all clear what has re-introduced this constraint, which was suggested in the code base a long time ago, but has not been observed on `main` at all recently.
* it is not clear whether this constraint is also linked to scale
investigate this constraint, and whether it applies to all jobs, or at some scale
implement rose-stem or launch-exe level testing that will fail for jobs that do not meet this constraint and would fail, with clear error messaging
Contributor guide
No contributing guide indexed for this repository
Research direction
Start by reading issues #470 and #519 and tracing where mpi_parts and mpi_parts_xios are validated or used in rose-stem and launch-exe for large weekly jobs. Reproduce jobs with divisible and non-divisible ratios at different scales, then add a test that captures the observed failure cases and reports a clear constraint error.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- fortran
- Domain
- hpc, testing
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100