#1426 closed Task (fixed)
ROBUST-01(2015WP)/MERCATOR-6 online coarsening
Reported by: | cbricaud | Owned by: | cbricaud |
---|---|---|---|
Priority: | normal | Milestone: | 2018 WP |
Component: | OCE | Version: | v3.6 |
Severity: | minor | Keywords: | 2015 v3.6 |
Cc: |
Commit History (0)
(No commits)
Change History (22)
comment:1 Changed 9 years ago by cbricaud
comment:2 Changed 9 years ago by cbricaud
https://forge.ipsl.jussieu.fr/nemo/changeset/5010
bug detected:
in crsfld: crs_volt_wgt,crs_surfu_wgt and crs_surfv_wgt are used to compute hdiv.
but they are declared/allocated, but no computed. do hdiv=0 in coarsened ouputs.
change computing of hdivn_crs(ji,jj,jk)
allow CRS to run with ln_nogather=T
comment:3 Changed 9 years ago by cbricaud
- Milestone set to 2015 nemo_v3_6_STABLE
comment:4 Changed 8 years ago by cbricaud
changes at rev5601:
https://forge.ipsl.jussieu.fr/nemo/changeset/5601
running with ORCA025_LIM2, time-splitting, fixed volume and age tracer
comment:5 Changed 8 years ago by cbricaud
merge change from trunk rev 5003 to 5519 ( rev where branche 3.6_stable were created ):
svn merge -r 5003:5519 svn+ssh://cbricaud@forge.ipsl.jussieu.fr/ipsl/forge/projets/nemo/svn/trunk .
commit at rev 5602.
https://forge.ipsl.jussieu.fr/nemo/changeset/5602
comment:6 Changed 7 years ago by nicolasmartin
- Keywords 2015 nemo_v3_6* added
comment:7 Changed 7 years ago by cbricaud
the branch containing the online coarsening of BCG model is available here:
https://forge.ipsl.jussieu.fr/nemo/browser/branches/2015/dev_r5003_MERCATOR6_CRS
This branch is associated nemo 3.6 stable ( available here: https://forge.ipsl.jussieu.fr/nemo/browser/branches/2015/nemo_v3_6_STABLE/NEMOGCM )
This branch is phased with rev 7213 of nemo 3.6 stable.
Coarsening capacity can be used:
- for output of physics variables at a lower resolution
- or to run a configuration using TOP online at a lower resolution
A "readme" is available in the branch : NEMOGCM/CRS_help/README_COARSENING.txt
At the moment, coarsening can not be activated with PISCES; it will be available in the next weeks.
Different runs have been performed with Age tracer to validate coarsening implementation:
- a run 'full resolution' based on eORCA025_LIM3 ( physic and passive tracer at 1/4° resolution )
- a run 'low resolution' based on eORCA075_LIM3 ( physic and passive tracer at 3/4° resolution )
- a run with coarsening based on eORCA025_LIM3 ( physic at 1/4° resolution and passive tracer at 3/4° resolution )
Age tracer from the coarsened run well reproduce age tracer from 'full resolution' run.
comment:8 Changed 7 years ago by cbricaud
At the revision 7345 of this branch is, PISCES routines has not been impacted.
comment:9 Changed 7 years ago by cbricaud
commit a first implementation of coarsening in PISCES at rev7398 ( https://forge.ipsl.jussieu.fr/nemo/changeset/7398 )
PISCES is running with coarsening ( not crash ) but there is a problem of CHL productions. problem under investigation
Note that, as TOP need to read data from files for tracers CI , or as PISCES needs too to read data from files , there is a change in IOM to tell it when reading is done in 'coarsened space'
comment:10 Changed 7 years ago by cbricaud
http://forge.ipsl.jussieu.fr/nemo/changeset/7520 : now PISCES is running with online coarsening
http://forge.ipsl.jussieu.fr/nemo/changeset/7521 : big correction for mother grid domain definition in coarsening
comment:11 Changed 6 years ago by cbricaud
Different bugfix have been committed :
see http://forge.ipsl.jussieu.fr/nemo/changeset?reponame=&new=8103%40branches%2F2015%2Fdev_r5003_MERCATOR6_CRS&old=7806%40branches%2F2015%2Fdev_r5003_MERCATOR6_CRS
- 2 bug fixes on lateral diffusion
- maskland option ok for coarsening variables
- horizontal surfaces changed fot isopycnal diffusion and surface fluxes coarsening
- bugfixes on coarsned grid local indices computing
- restartabity ok
comment:12 Changed 6 years ago by nicolasmartin
- Owner changed from nemo to cbricaud
- Status changed from new to assigned
comment:13 Changed 6 years ago by nemo
- Type changed from Development to Task
Remove 'Development' type
comment:14 Changed 6 years ago by nicolasmartin
- Milestone 2015 nemo_v3_6_STABLE deleted
- workplan_status set to unknown
comment:15 Changed 6 years ago by nemo
- Keywords release-3.6* added; nemo_v3_6* removed
comment:16 Changed 6 years ago by nemo
- Keywords release-3.6* removed
comment:17 Changed 5 years ago by nemo
- Milestone set to Unscheduled
Batch modification of milestone field set to 'Unscheduled'
comment:18 Changed 5 years ago by nicolasmartin
- Milestone changed from Unscheduled to 2018 WP
- Summary changed from ROBUST-1 /MERCATOR-6 online coarsening to ROBUST-1 /MERCATOR-6 online coarsening
Now the CRS branch is located source:/NEMO/branches/2018/dev_r5003_MERCATOR6_CRS
Should consider an last one phase with 3.6 release at some point.
comment:19 Changed 5 years ago by nicolasmartin
- Summary changed from ROBUST-1 /MERCATOR-6 online coarsening to ROBUST-1(2015WP)/MERCATOR-6 online coarsening
comment:20 Changed 5 years ago by nicolasmartin
- Summary changed from ROBUST-1(2015WP)/MERCATOR-6 online coarsening to ROBUST-01(2015WP)/MERCATOR-6 online coarsening
comment:21 Changed 5 years ago by gsamson
- Resolution set to fixed
- Status changed from assigned to closed
comment:22 Changed 20 months ago by nemo
- Keywords v3.6 added
https://forge.ipsl.jussieu.fr/nemo/changeset/5007
in crs_dom_def: jpj_crs can be increased. all arrays with jpj_crs as dimension should be allocated affter. so crs_dom_alloc is splitted and crs_dom_alloc1 is called after dimension computing
in crs_dom_def: noso is refering to the proc on which where are computing, but it used in the loop on processors jn.
so we define iprocso which refers to proc jn.
in crs_dom_def: we change the loops on jn which are not compatible with mpp_ini2.
results are reproductible for 24.20.480 with the old loop and the new loop.
results are reproductible with the new loop for 24.20.480 and 24.20.390 .
crsini.F90: avoid divsion per 0. for facsurfu/v.
crsfld.F90: avoid division pr 0. with crs_volt_wgt for hdivn_crs
mppini.F90 and mppini_2.h90: modifications to allow landprocessor removing with CRS.