Changeset 15043 for NEMO/branches/2021/ticket2696_icb_halo1_halo2_compatibility/src/OCE/ICB/icbutl.F90

Timestamp:

2021-06-22T13:26:16+02:00 (3 years ago)

Author:

acc

Message:

#2696 .A possible fix for obtaining identical results with icebergs active and with either
nn_hls=1 or nn_hls=2. This solution works by ensuring the real-valued iceberg position
calculations are done with identical numbers in both cases. Iceberg positions are stored
as a continuous local grid-cell value and a factor of (nn_hls-1) is subtracted from
initial positions to ensure equivalence in both cases. This factor needs to be added
back whenever iceberg positions are converted to local indices. The changes in this
commit are enough to ensure a full ORCA2_ICE_PISCES SETTE test is passed in both case
(992 timesteps) and provides identical results. Suitability of this solution for inclusion
in the trunk needs to be discussed and verified on other platforms and compilers.

File:

• NEMO/branches/2021/ticket2696_icb_halo1_halo2_compatibility/src/OCE/ICB/icbutl.F90 (modified) (2 diffs)

NEMO/branches/2021/ticket2696_icb_halo1_halo2_compatibility/src/OCE/ICB/icbutl.F90

@@ -300 +300 @@
          zwj = pj - 0.5_wp - REAL(kij,wp)
       END SELECT
+      kii = kii + (nn_hls-1)
+      kij = kij + (nn_hls-1)
       !
       ! compute weight
@@ -461 +463 @@
 
       ! conversion to local domain (no need to do a sanity check already done in icbpos)
-      ii = mi1(ii)
-      ij = mj1(ij)
+      ii = mi1(ii) + (nn_hls-1)
+      ij = mj1(ij) + (nn_hls-1)
       !
       IF(    0.0_wp <= zi .AND. zi < 0.5_wp   ) THEN