Opened 7 years ago

Closed 7 years ago

#1440 closed Enhancement (fixed)

use of wrk_alloc( jpidta, jpjdta...

Reported by: smasson Owned by: nemo
Priority: normal Milestone:
Component: OCE Version: trunk
Severity: Keywords:
Cc: Branch review:
MP ready?: Task progress:


There is a few routines of the code that use work_alloc for arrays that have the size of the full domain (jpidta, jpjdta) instead of the size of the local domain (jpi,jpj). I let you imagine the memory print of this when running ORCA12 on 10000 cores for example…

Clearly using (jpidta, jpjdta) should be avoided but when it is really necessary, we should not use work_alloc as the whole idea of work_alloc/work_dealloc is to keep the array allocated in memory. So you create huge arrays just for initialization phase and you keep them in memory during all your run!

This concerns 

OPA_SRC/DOM/domzgr.F90 : allocate the arrays in all cases but use them only in cases with flat or analytical bathymetry!

OPA_SRC/LDF/ldfdyn_c2d.h90 : only for ORCA2 and ORCA1 and will be modified in a future release → no big deal

OPA_SRC/C1D/domc1d.F90 : in this case jpidta = jpjdta = 3, so it is ok 

Commit History (1)


supress wrk_alloc( jpidta, jpjdta, see ticket #1440

Change History (1)

comment:1 Changed 7 years ago by smasson

  • Resolution set to fixed
  • Status changed from new to closed

fixed, see changeset r5040

Note: See TracTickets for help on using tickets.