[Mitgcm-support] Re: verification/hs94.cs-32x32x5

[mitgcm-support at dev.mitgcm.org]

Chris Hill wrote:
> Works better now. Had prob. with 750 sec deltat NaN at
> 11.1 days. Cut timestep to 300s and things seemed
> better.

720 seconds ran fine for me all night. You have a bug.

The ini_depths.F bug is related. I've fixed the problem in
ini_masks_etc.F instead (cvs update ini_masks_etc.F):

diff -r1.20.2.2 ini_masks_etc.F
180,182c180,183
<       _EXCH_XYZ_R4(hFacW , myThid )
<       _EXCH_XYZ_R4(hFacS , myThid )
< C     Re-do hFacW and hFacS (at U and V points)
---
>       CALL EXCH_UV_XYZ_RL(hFacW,hFacS,.FALSE.,myThid)
> C     The following block allows thin walls representation of non-periodic
> C     boundaries such as happen on the lat-lon grid at the N/S poles.
> C     We should really supply a flag for doing this.
187,195c188,190
<           DO I=1-Olx+1,sNx+Olx ! Note range
<            hFacW(I,J,K,bi,bj)=
<      &       MIN(hFacC(I,J,K,bi,bj),hFacC(I-1,J,K,bi,bj))
<           ENDDO
<          ENDDO
<          DO I=1-Olx,sNx+Olx
<           DO J=1-Oly+1,sNy+Oly ! Note range
<            hFacS(I,J,K,bi,bj)=
<      &       MIN(hFacC(I,J,K,bi,bj),hFacC(I,J-1,K,bi,bj))
---
>           DO I=1-Olx,sNx+Olx
>            IF (DYG(I,J,bi,bj).EQ.0.) hFacW(I,J,K,bi,bj)=0.
>            IF (DXG(I,J,bi,bj).EQ.0.) hFacS(I,J,K,bi,bj)=0.