Misplaced update of time varying lateral diffusion for time splitting

[jchanut]

Change step order to compute time varying (in direction or magnitude) lateral viscosities/diffusivities before the update of momentum trend.
No error with iso-level momentum diffusion and iso-level/isopycnal tracer diffusion.

Apart from this, I have some trouble to understand how diffusion in the horizontal direction is done in s-coordinates. Even Gurvan does not seem to know...see below in particular the conditionnal expression in ldf_slp_init;-) Can anyone explain ?

SUBROUTINE ldf_slp_init

         !!gm I no longer understand this.....
         IF( (ln_traldf_hor .OR. ln_dynldf_hor) .AND. .NOT. (lk_vvl .AND. ln_rstart) ) THEN
            IF(lwp)   WRITE(numout,*) '          Horizontal mixing in s-coordinate: slope = slope of s-surfaces'
            DO jk = 1, jpk
               DO jj = 2, jpjm1
                  DO ji = fs_2, fs_jpim1   ! vector opt.
                     uslp (ji,jj,jk) = -1./e1u(ji,jj) * ( fsdept(ji+1,jj,jk) - fsdept(ji ,jj ,jk) ) * umask(ji,jj,jk)
                     vslp (ji,jj,jk) = -1./e2v(ji,jj) * ( fsdept(ji,jj+1,jk) - fsdept(ji ,jj ,jk) ) * vmask(ji,jj,jk)
                     wslpi(ji,jj,jk) = -1./e1t(ji,jj) * ( fsdepw(ji+1,jj,jk) - fsdepw(ji-1,jj,jk) ) * tmask(ji,jj,jk) * 0.5
                     wslpj(ji,jj,jk) = -1./e2t(ji,jj) * ( fsdepw(ji,jj+1,jk) - fsdepw(ji,jj-1,jk) ) * tmask(ji,jj,jk) * 0.5
                  END DO
               END DO
            END DO
!...
!...
END SUBROUTINE ldf_slp_init

[jchanut]

Changed time step order in revision 4369