Opened 8 years ago
Closed 7 years ago
#1628 closed Bug (fixed)
bug when key_mpp_mpi is not active
Reported by: | clem | Owned by: | nemo |
---|---|---|---|
Priority: | low | Milestone: | |
Component: | OCE | Version: | v3.6 |
Severity: | Keywords: | ||
Cc: |
Description
Multiple communications (lbc_lnk_multi) have not been coded when key_mpp_mpi is inactive. Since these communications are present in dynspg_ts and limrhg, the model crashes when running orca2-lim3 (for example).
Therefore I propose to modify lbclnk.F90 (as shown in the attached file). It looks to me the easiest way to deal with this issue but additional expertise would be appreciated
Commit History (0)
(No commits)
Attachments (1)
Change History (3)
Changed 8 years ago by clem
comment:1 Changed 8 years ago by mcastril
comment:2 Changed 7 years ago by timgraham
- Resolution set to fixed
- Status changed from new to closed
This has already been fixed when dev_r5546_CNRS19_HPC_scalability was merged into 3.6 STABLE
Note: See
TracTickets for help on using
tickets.
I think we have 2 open tickets of these. We developed an alternative version, without load_array, and only calling to lbc_lnk single from the lbc_lnk_2d_9. We only created lbc_lnk_2d_multiple because is called directly from limhdf in the new optimization, but in fact it only does a loop with the different variables.
Maybe your approach is more elegant, but in fact in serial mode it has any benefit to perform the packing, and our code is a little bit simpler (and with less number of lines).