Profiling UKESM0.4 vs UKESM0.5

UKESM0,4 is u-ak928 (a copy of u-ai432) and UKESM0.5 is u-ak929 (a copy of u-aj599) and they're all

  • Run for one month
  • (ATM_PROCX,ATM_PROCY)=(48,24)
  • OMPTHR_ATM=2
  • IOS_NPROC=6
  • (NEMO_IPROC,NEMO_JPROC)=(12,9)
  • (CICE_BLKX,CICE_BLKY)=(30,37)
  • total nodes=68

Profiling the top routines

UKESM0.4

Routines
UM_SHELL (1,512s)
U_MODEL_4A (1,502s)
ATM_STEP_4A* (1,324s) OASIS3_ PUTA2O (7s) OASIS3_ GETO2A (57s)
ATMOS _PHYS- ICS1 (451s) ATMOS _PHYS- ICS2 (85s) EG_ SL_ HELM- HOLTZ (121s) TR_ SET_ PHYS _4A* (29s) EG_CORRECT _TRACERS _PRIESTLEY (29s) SL_ TRAC- ER1_ 4A (49s) EG_ SL_ MOI- STURE (25s) EG_SL_ FULL_WIND (49s)  ⇓  UKCA_MAIN1 (176s) ICE_ SHEET_ MASS (5s) OASIS3_ GET (57s)
Profile for ATMOS_ PHYSICS1 Profile for ATMOS_ PHYSICS2 and EG_SL_ HELMHOLTZ EG_SL_WIND_U, EG_SL_WIND_V & EG_SL_WIND_W (8 + 7 + 16 = 31s) STASH (128s) See profiling for UKCA_ MAIN1 GLOBAL_ 2D_ SUMS** (47s)
Itself (10s) EG_INTERPOLATION _ETA_PMF (50s) DEP- ARTURE_ POINT _ETA (32s) STWORK (127s)
EG_INTERPOLATION _ETA (50s) Itself (3s) PP_ HEAD (50s) EXP- PXI (41s, itself)
EG_ CUBIC_ LAG- RANGE (11s, itself) MONO_ ENFORCE (4s, itself) Itself (30s)
*should also link to SWAP_BOUNDS, like many other returns.
** GLOBAL_2D_SUMS is called by other routines as well

UKESM0.5

Routines
UM_SHELL (1,753s)
U_MODEL_4A (1,741s)
ATM_STEP_4A* (1,325s) OASIS3_ PUTA2O (3s) OASIS3_ GETO2A (286s)
ATMOS _PHYS- ICS1 (449s) ATMOS _PHYS- ICS2 (87s) EG_ SL_ HELM- HOLTZ (122s) TR_ SET_ PHYS _4A* (29s) EG_CORRECT _TRACERS _PRIESTLEY (29s) SL_ TRAC- ER1_ 4A (49s) EG_ SL_ MOI- STURE (122s) EG_SL_ FULL_WIND (48s)  ⇓  UKCA_MAIN1 (179s) ICE_ SHEET_ MASS (2s) OASIS3_ GET (286s)
Profile for ATMOS_ PHYSICS1 Profile for ATMOS_ PHYSICS2 and EG_SL_ HELMHOLTZ EG_SL_WIND_U, EG_SL_WIND_V & EG_SL_WIND_W (8 + 7 + 15 = 31s) STASH (129s) See profiling for UKCA_ MAIN1 GLOBAL_ 2D_ SUMS** (41s)
Itself (10s) EG_INTERPOLATION _ETA_PMF (49s) DEP- ARTURE_ POINT _ETA (31s) STWORK (129s)
EG_INTERPOLATION _ETA (49s) Itself (3s) PP_ HEAD (52s) EXP- PXI (42s, itself)
EG_ CUBIC_ LAG- RANGE (11s, itself) MONO_ ENFORCE (4s, itself) Itself (29s)
*should also link to SWAP_BOUNDS, like many other returns.
** GLOBAL_2D_SUMS is called by other routines as well

It looks like the extra time comes from waiting for the ocean.